Package 'timereg'

Title: Flexible Regression Models for Survival Data
Description: Programs for Martinussen and Scheike (2006), `Dynamic Regression Models for Survival Data', Springer Verlag. Plus more recent developments. Additive survival model, semiparametric proportional odds model, fast cumulative residuals, excess risk models and more. Flexible competing risks regression including GOF-tests. Two-stage frailty modelling. PLS for the additive risk model. Lasso in the 'ahaz' package.
Authors: Thomas Scheike [aut, cre], Torben Martinussen [aut], Jeremy Silver [ctb], Klaus K. Holst [ctb]
Maintainer: Thomas Scheike <[email protected]>
License: GPL (>= 2)
Version: 2.0.6
Built: 2024-11-05 04:57:38 UTC
Source: https://github.com/scheike/timereg

Help Index

Fit additive hazards model

Description

Fits both the additive hazards model of Aalen and the semi-parametric additive hazards model of McKeague and Sasieni. Estimates are un-weighted. Time dependent variables and counting process data (multiple events per subject) are possible.

Usage

aalen(
  formula = formula(data),
  data = parent.frame(),
  start.time = 0,
  max.time = NULL,
  robust = 1,
  id = NULL,
  clusters = NULL,
  residuals = 0,
  n.sim = 1000,
  weighted.test = 0,
  covariance = 0,
  resample.iid = 0,
  deltaweight = 1,
  silent = 1,
  weights = NULL,
  max.clust = 1000,
  gamma = NULL,
  offsets = 0,
  caseweight = NULL
)

Arguments

formula

a formula object with the response on the left of a '~' operator, and the independent terms on the right as regressors.The response must be a survival object as returned by the ‘Surv’ function. Time- invariant regressors are specified by the wrapper const(), and cluster variables (for computing robust variances) by the wrapper cluster().
data

a data.frame with the variables.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. Default is max of data.
robust

to compute robust variances and construct processes for resampling. May be set to 0 to save memory.
id

For timevarying covariates the variable must associate each record with the id of a subject.
clusters

cluster variable for computation of robust variances.
residuals

to returns residuals that can be used for model validation in the function cum.residuals
n.sim

number of simulations in resampling.
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
covariance

to compute covariance estimates for nonparametric terms rather than just the variances.
resample.iid

to return i.i.d. representation for nonparametric and parametric terms.
deltaweight

uses weights to estimate semiparametric model, under construction, default=1 is standard least squares estimates
silent

set to 0 to print warnings for non-inverible design-matrices for different timepoints, default is 1.
weights

weights for estimating equations.
max.clust

sets the total number of i.i.d. terms in i.i.d. decompostition. This can limit the amount of memory used by coarsening the clusters. When NULL then all clusters are used. Default is 1000 to save memory and time.
gamma

fixes gamme at this value for estimation.
offsets

offsets for the additive model, to make excess risk modelling.
caseweight

caseweight: mutiplied onto dN for score equations.

Details

Resampling is used for computing p-values for tests of time-varying effects.

The modelling formula uses the standard survival modelling given in the survival package.

The data for a subject is presented as multiple rows or 'observations', each of which applies to an interval of observation (start, stop]. For counting process data with the )start,stop] notation is used, the 'id' variable is needed to identify the records for each subject. The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Value

returns an object of type "aalen". With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.

var.cum

the martingale based pointwise variance estimates for cumulatives.
robvar.cum

robust pointwise variances estimates for cumulatives.
gamma

estimate of parametric components of model.

var.gamma

variance for gamma.

robvar.gamma

robust variance for gamma.

residuals

list with residuals. Estimated martingale increments (dM) and corresponding time vector (time).
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.

test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect over time.

sim.test.procBeqC

list of 50 random realizations of test-processes under null based on resampling.
covariance

covariances for nonparametric terms of model.
B.iid

Resample processes for nonparametric terms of model.
gamma.iid

Resample processes for parametric terms of model.
deviance

Least squares of increments.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

data(sTRACE)
# Fits Aalen model 
out<-aalen(Surv(time,status==9)~age+sex+diabetes+chf+vf,
sTRACE,max.time=7,n.sim=100)

summary(out)
par(mfrow=c(2,3))
plot(out)

# Fits semi-parametric additive hazards model 
out<-aalen(Surv(time,status==9)~const(age)+const(sex)+const(diabetes)+chf+vf,
sTRACE,max.time=7,n.sim=100)

summary(out)
par(mfrow=c(2,3))
plot(out)

## Excess risk additive modelling 
data(mela.pop)
dummy<-rnorm(nrow(mela.pop));

# Fits Aalen model  with offsets 
out<-aalen(Surv(start,stop,status==1)~age+sex+const(dummy),
mela.pop,max.time=7,n.sim=100,offsets=mela.pop$rate,id=mela.pop$id,
gamma=0)
summary(out)
par(mfrow=c(2,3))
plot(out,main="Additive excess riks model")

# Fits semi-parametric additive hazards model  with offsets 
out<-aalen(Surv(start,stop,status==1)~age+const(sex),
mela.pop,max.time=7,n.sim=100,offsets=mela.pop$rate,id=mela.pop$id)
summary(out)
plot(out,main="Additive excess riks model")

The Bone Marrow Transplant Data

Description

Bone marrow transplant data with 408 rows and 5 columns.

Format

The data has 408 rows and 5 columns.

cause

a numeric vector code. Survival status. 1: dead from treatment related causes, 2: relapse , 0: censored.

time

a numeric vector. Survival time.

platelet

a numeric vector code. Plalelet 1: more than 100 x 10910^9 per L, 0: less.

tcell

a numeric vector. T-cell depleted BMT 1:yes, 0:no.

age

a numeric vector code. Age of patient, scaled and centered ((age-35)/15).

Source

Simulated data

References

NN

Examples

data(bmt)
names(bmt)

The multicenter AIDS cohort study

Description

CD4 counts collected over time.

Format

This data frame contains the following columns:

obs

a numeric vector. Number of observations.

id

a numeric vector. Id of subject.

visit

a numeric vector. Timings of the visits in years.

smoke

a numeric vector code. 0: non-smoker, 1: smoker.

age

a numeric vector. Age of the patient at the start of the trial.

cd4

a numeric vector. CD4 percentage at the current visit.

cd4.prev

a numeric vector. CD4 level at the preceding visit.

precd4

a numeric vector. Post-infection CD4 percentage.

lt

a numeric vector. Gives the starting time for the time-intervals.

rt

a numeric vector. Gives the stopping time for the time-interval.

Source

MACS Public Use Data Set Release PO4 (1984-1991). See reference.

References

Kaslow et al. (1987), The multicenter AIDS cohort study: rational, organisation and selected characteristics of the participants. Am. J. Epidemiology 126, 310–318.

Examples

data(cd4)
names(cd4)

Competings Risks Regression

Description

Fits a semiparametric model for the cause-specific quantities :

P(T<t,cause=1∣x,z)=P1(t,x,z)=h(g(t,x,z))P(T < t, cause=1 | x,z) = P_1(t,x,z) = h( g(t,x,z) )

for a known link-function h()h() and known prediction-function g(t,x,z)g(t,x,z) for the probability of dying from cause 1 in a situation with competing causes of death.

Usage

comp.risk(
  formula,
  data = parent.frame(),
  cause,
  times = NULL,
  Nit = 50,
  clusters = NULL,
  est = NULL,
  fix.gamma = 0,
  gamma = 0,
  n.sim = 0,
  weighted = 0,
  model = "fg",
  detail = 0,
  interval = 0.01,
  resample.iid = 1,
  cens.model = "KM",
  cens.formula = NULL,
  time.pow = NULL,
  time.pow.test = NULL,
  silent = 1,
  conv = 1e-06,
  weights = NULL,
  max.clust = 1000,
  n.times = 50,
  first.time.p = 0.05,
  estimator = 1,
  trunc.p = NULL,
  cens.weights = NULL,
  admin.cens = NULL,
  conservative = 1,
  monotone = 0,
  step = NULL
)

Arguments

formula

a formula object, with the response on the left of a '~' operator, and the terms on the right. The response must be a survival object as returned by the ‘Event’ function. The status indicator is not important here. Time-invariant regressors are specified by the wrapper const(), and cluster variables (for computing robust variances) by the wrapper cluster().
data

a data.frame with the variables.
cause

For competing risk models specificies which cause we consider.
times

specifies the times at which the estimator is considered. Defaults to all the times where an event of interest occurs, with the first 10 percent or max 20 jump points removed for numerical stability in simulations.
Nit

number of iterations for Newton-Raphson algorithm.
clusters

specifies cluster structure, for backwards compability.
est

possible starting value for nonparametric component of model.
fix.gamma

to keep gamma fixed, possibly at 0.
gamma

starting value for constant effects.
n.sim

number of simulations in resampling.
weighted

Not implemented. To compute a variance weighted version of the test-processes used for testing time-varying effects.
model

"additive", "prop"ortional, "rcif", or "logistic".
detail

if 0 no details are printed during iterations, if 1 details are given.
interval

specifies that we only consider timepoints where the Kaplan-Meier of the censoring distribution is larger than this value.
resample.iid

to return the iid decomposition, that can be used to construct confidence bands for predictions
cens.model

specified which model to use for the ICPW, KM is Kaplan-Meier alternatively it may be "cox"
cens.formula

specifies the regression terms used for the regression model for chosen regression model. When cens.model is specified, the default is to use the same design as specified for the competing risks model.
time.pow

specifies that the power at which the time-arguments is transformed, for each of the arguments of the const() terms, default is 1 for the additive model and 0 for the proportional model.
time.pow.test

specifies that the power the time-arguments is transformed for each of the arguments of the non-const() terms. This is relevant for testing if a coefficient function is consistent with the specified form A_l(t)=beta_l t^time.pow.test(l). Default is 1 for the additive model and 0 for the proportional model.
silent

if 0 information on convergence problems due to non-invertible derviates of scores are printed.
conv

gives convergence criterie in terms of sum of absolute change of parameters of model
weights

weights for estimating equations.
max.clust

sets the total number of i.i.d. terms in i.i.d. decompostition. This can limit the amount of memory used by coarsening the clusters. When NULL then all clusters are used. Default is 1000 to save memory and time.
n.times

only uses 50 points for estimation, if NULL then uses all points, subject to p.start condition.
first.time.p

first point for estimation is pth percentile of cause jump times.
estimator

default estimator is 1.
trunc.p

truncation weight for delayed entry, P(T > entry.time | Z_i), typically Cox model.
cens.weights

censoring weights can be given here rather than calculated using the KM, cox or aalen models.
admin.cens

censoring times for the administrative censoring
conservative

set to 0 to compute correct variances based on censoring weights, default is conservative estimates that are much quicker.
monotone

monotone=0, uses estimating equations

(DβP1)w(t)(Y(t)/Gc(t)−P1(t,X))(D_\beta P_1) w(t) ( Y(t)/G_c(t) - P_1(t,X))

montone=1 uses

w(t)X(Y(t)/Gc(t)−P1(t,X))w(t) X ( Y(t)/G_c(t) - P_1(t,X))

step

step size for Fisher-Scoring algorithm.

Details

We consider the following models : 1) the additive model where h(x)=1−exp⁡(−x)h(x)=1-\exp(-x) and

g(t,x,z)=xTA(t)+(diag(tp)z)Tβg(t,x,z) = x^T A(t) + (diag(t^p) z)^T \beta

2) the proportional setting that includes the Fine & Gray (FG) "prop" model and some extensions where h(x)=1−exp⁡(−exp⁡(x))h(x)=1-\exp(-\exp(x)) and

g(t,x,z)=(xTA(t)+(diag(tp)z)Tβ)g(t,x,z) = (x^T A(t) + (diag(t^p) z)^T \beta)

The FG model is obtained when x=1x=1, but the baseline is parametrized as exp⁡(A(t))\exp(A(t)).

The "fg" model is a different parametrization that contains the FG model, where h(x)=1−exp⁡(−x)h(x)=1-\exp(-x) and

g(t,x,z)=(xTA(t))exp⁡((diag(tp)z)Tβ)g(t,x,z) = (x^T A(t)) \exp((diag(t^p) z)^T \beta)

The FG model is obtained when x=1x=1.

3) a "logistic" model where h(x)=exp⁡(x)/(1+exp⁡(x))h(x)=\exp(x)/( 1+\exp(x)) and

g(t,x,z)=xTA(t)+(diag(tp)z)Tβg(t,x,z) = x^T A(t) + (diag(t^p) z)^T \beta

The "logistic2" is

P1(t,x,z)=xTA(t)exp((diag(tp)z)Tβ)/(1+xTA(t)exp((diag(tp)z)Tβ))P_1(t,x,z) = x^T A(t) exp((diag(t^p) z)^T \beta)/ (1+ x^T A(t) exp((diag(t^p) z)^T \beta))

The simple logistic model with just a baseline can also be fitted by an alternative procedure that has better small sample properties see prop.odds.subist().

4) the relative cumulative incidence function "rcif" model where h(x)=exp⁡(x)h(x)=\exp(x) and

g(t,x,z)=xTA(t)+(diag(tp)z)Tβg(t,x,z) = x^T A(t) + (diag(t^p) z)^T \beta

The "rcif2"

P1(t,x,z)=(xTA(t))exp⁡((diag(tp)z)Tβ)P_1(t,x,z) = (x^T A(t)) \exp((diag(t^p) z)^T \beta)

Where p by default is 1 for the additive model and 0 for the other models. In general p may be powers of the same length as z.

Since timereg version 1.8.4. the response must be specified with the Event function instead of the Surv function and the arguments. For example, if the old code was

comp.risk(Surv(time,cause>0)~x1+x2,data=mydata,cause=mydata$cause,causeS=1)

the new code is

comp.risk(Event(time,cause)~x1+x2,data=mydata,cause=1)

Also the argument cens.code is now obsolete since cens.code is an argument of Event.

Value

returns an object of type 'comprisk'. With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

pointwise variances estimates.

gamma

estimate of proportional odds parameters of model.
var.gamma

variance for gamma.

score

sum of absolute value of scores.
gamma2

estimate of constant effects based on the non-parametric estimate. Used for testing of constant effects.
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
conf.band

resampling based constant to construct 95% uniform confidence bands.
B.iid

list of iid decomposition of non-parametric effects.
gamma.iid

matrix of iid decomposition of parametric effects.
test.procBeqC

observed test process for testing of time-varying effects
sim.test.procBeqC

50 resample processes for for testing of time-varying effects
conv

information on convergence for time points used for estimation.

Author(s)

Thomas Scheike

References

Scheike, Zhang and Gerds (2008), Predicting cumulative incidence probability by direct binomial regression,Biometrika, 95, 205-220.

Scheike and Zhang (2007), Flexible competing risks regression modelling and goodness of fit, LIDA, 14, 464-483.

Martinussen and Scheike (2006), Dynamic regression models for survival data, Springer.

Examples

data(bmt); 

clust <- rep(1:204,each=2)
addclust<-comp.risk(Event(time,cause)~platelet+age+tcell+cluster(clust),data=bmt,
cause=1,resample.iid=1,n.sim=100,model="additive")
###

addclust<-comp.risk(Event(time,cause)~+1+cluster(clust),data=bmt,cause=1,
		    resample.iid=1,n.sim=100,model="additive")
pad <- predict(addclust,X=1)
plot(pad)

add<-comp.risk(Event(time,cause)~platelet+age+tcell,data=bmt,
cause=1,resample.iid=1,n.sim=100,model="additive")
summary(add)

par(mfrow=c(2,4))
plot(add); 
### plot(add,score=1) ### to plot score functions for test

ndata<-data.frame(platelet=c(1,0,0),age=c(0,1,0),tcell=c(0,0,1))
par(mfrow=c(2,3))
out<-predict(add,ndata,uniform=1,n.sim=100)
par(mfrow=c(2,2))
plot(out,multiple=0,uniform=1,col=1:3,lty=1,se=1)

## fits additive model with some constant effects 
add.sem<-comp.risk(Event(time,cause)~
const(platelet)+const(age)+const(tcell),data=bmt,
cause=1,resample.iid=1,n.sim=100,model="additive")
summary(add.sem)

out<-predict(add.sem,ndata,uniform=1,n.sim=100)
par(mfrow=c(2,2))
plot(out,multiple=0,uniform=1,col=1:3,lty=1,se=0)

## Fine & Gray model 
fg<-comp.risk(Event(time,cause)~
const(platelet)+const(age)+const(tcell),data=bmt,
cause=1,resample.iid=1,model="fg",n.sim=100)
summary(fg)

out<-predict(fg,ndata,uniform=1,n.sim=100)

par(mfrow=c(2,2))
plot(out,multiple=1,uniform=0,col=1:3,lty=1,se=0)

## extended model with time-varying effects
fg.npar<-comp.risk(Event(time,cause)~platelet+age+const(tcell),
data=bmt,cause=1,resample.iid=1,model="prop",n.sim=100)
summary(fg.npar); 

out<-predict(fg.npar,ndata,uniform=1,n.sim=100)
head(out$P1[,1:5]); head(out$se.P1[,1:5])

par(mfrow=c(2,2))
plot(out,multiple=1,uniform=0,col=1:3,lty=1,se=0)

## Fine & Gray model with alternative parametrization for baseline
fg2<-comp.risk(Event(time,cause)~const(platelet)+const(age)+const(tcell),data=bmt,
cause=1,resample.iid=1,model="prop",n.sim=100)
summary(fg2)

#################################################################
## Delayed entry models, 
#################################################################
nn <- nrow(bmt)
entrytime <- rbinom(nn,1,0.5)*(bmt$time*runif(nn))
bmt$entrytime <- entrytime
times <- seq(5,70,by=1)

bmtw <- prep.comp.risk(bmt,times=times,time="time",entrytime="entrytime",cause="cause")

## non-parametric model 
outnp <- comp.risk(Event(time,cause)~tcell+platelet+const(age),
		   data=bmtw,cause=1,fix.gamma=1,gamma=0,
 cens.weights=bmtw$cw,weights=bmtw$weights,times=times,n.sim=0)
par(mfrow=c(2,2))
plot(outnp)

outnp <- comp.risk(Event(time,cause)~tcell+platelet,
		   data=bmtw,cause=1,
 cens.weights=bmtw$cw,weights=bmtw$weights,times=times,n.sim=0)
par(mfrow=c(2,2))
plot(outnp)


## semiparametric model 
out <- comp.risk(Event(time,cause)~const(tcell)+const(platelet),data=bmtw,cause=1,
 cens.weights=bmtw$cw,weights=bmtw$weights,times=times,n.sim=0)
summary(out)

Identifies parametric terms of model

Description

Specifies which of the regressors that have constant effect.

Usage

const(x)

Arguments

x

variable

Author(s)

Thomas Scheike

Identifies proportional excess terms of model

Description

Specifies which of the regressors that lead to proportional excess hazard

Usage

cox(x)

Arguments

x

variable

Author(s)

Thomas Scheike

Fit Cox-Aalen survival model

Description

Fits an Cox-Aalen survival model. Time dependent variables and counting process data (multiple events per subject) are possible.

Usage

cox.aalen(
  formula = formula(data),
  data = parent.frame(),
  beta = NULL,
  Nit = 20,
  detail = 0,
  start.time = 0,
  max.time = NULL,
  id = NULL,
  clusters = NULL,
  n.sim = 500,
  residuals = 0,
  robust = 1,
  weighted.test = 0,
  covariance = 0,
  resample.iid = 1,
  weights = NULL,
  rate.sim = 1,
  beta.fixed = 0,
  max.clust = 1000,
  exact.deriv = 1,
  silent = 1,
  max.timepoint.sim = 100,
  basesim = 0,
  offsets = NULL,
  strata = NULL,
  propodds = 0,
  caseweight = NULL
)

Arguments

formula

a formula object with the response on the left of a '~' operator, and the independent terms on the right as regressors. The response must be a survival object as returned by the ‘Surv’ function. Terms with a proportional effect are specified by the wrapper prop(), and cluster variables (for computing robust variances) by the wrapper cluster().
data

a data.frame with the variables.
beta

starting value for relative risk estimates.
Nit

number of iterations for Newton-Raphson algorithm.
detail

if 0 no details is printed during iterations, if 1 details are given.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. Default is max of data.
id

For timevarying covariates the variable must associate each record with the id of a subject.
clusters

cluster variable for computation of robust variances.
n.sim

number of simulations in resampling.
residuals

to returns residuals that can be used for model validation in the function cum.residuals. Estimated martingale increments (dM) and corresponding time vector (time). When rate.sim=1 returns estimated martingales, dM_i(t) and if rate.sim=0, returns a matrix of dN_i(t).
robust

to compute robust variances and construct processes for resampling. May be set to 0 to save memory and time, in particular for rate.sim=1.
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
covariance

to compute covariance estimates for nonparametric terms rather than just the variances.
resample.iid

to return i.i.d. representation for nonparametric and parametric terms. based on counting process or martingale resduals (rate.sim).
weights

weights for weighted analysis.
rate.sim

rate.sim=1 such that resampling of residuals is based on estimated martingales and thus valid in rate case, rate.sim=0 means that resampling is based on counting processes and thus only valid in intensity case.
beta.fixed

option for computing score process for fixed relative risk parameter
max.clust

sets the total number of i.i.d. terms in i.i.d. decompostition. This can limit the amount of memory used by coarsening the clusters. When NULL then all clusters are used. Default is 1000 to save memory and time.
exact.deriv

if 1 then uses exact derivative in last iteration, if 2 then uses exact derivate for all iterations, and if 0 then uses approximation for all computations and there may be a small bias in the variance estimates. For Cox model always exact and all options give same results.
silent

if 1 then opppresses some output.
max.timepoint.sim

considers only this resolution on the time scale for simulations, see time.sim.resolution argument
basesim

1 to get simulations for cumulative baseline, including tests for contant effects.
offsets

offsets for analysis on log-scale. RR=exp(offsets+ x beta).
strata

future option for making strata in a different day than through X design in cox-aalen model (~-1+factor(strata)).
propodds

if 1 will fit the proportional odds model. Slightly less efficient than prop.odds() function but much quicker, for large data this also works.
caseweight

these weights have length equal to number of jump times, and are multiplied all jump times dN. Useful for getting the program to fit for example the proportional odds model or frailty models.

Details

λi(t)=Yi(t)(XiT(t)α(t))exp⁡(ZiTβ)\lambda_{i}(t) = Y_i(t) ( X_{i}^T(t) \alpha(t) ) \exp(Z_{i}^T \beta )

The model thus contains the Cox's regression model as special case.

To fit a stratified Cox model it is important to parametrize the baseline apppropriately (see example below).

Resampling is used for computing p-values for tests of time-varying effects. Test for proportionality is considered by considering the score processes for the proportional effects of model.

The modelling formula uses the standard survival modelling given in the survival package.

The data for a subject is presented as multiple rows or 'observations', each of which applies to an interval of observation (start, stop]. For counting process data with the )start,stop] notation is used, the 'id' variable is needed to identify the records for each subject. The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Value

returns an object of type "cox.aalen". With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.

var.cum

the martingale based pointwise variance estimates.

robvar.cum

robust pointwise variances estimates.

gamma

estimate of parametric components of model.

var.gamma

variance for gamma sandwhich estimator based on optional variation estimator of score and 2nd derivative.
robvar.gamma

robust variance for gamma.

residuals

list with residuals.
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.

test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect over time.

sim.test.procBeqC

list of 50 random realizations of test-processes under null based on resampling.
covariance

covariances for nonparametric terms of model.
B.iid

Resample processes for nonparametric terms of model.
gamma.iid

Resample processes for parametric terms of model.
loglike

approximate log-likelihood for model, similar to Cox's partial likelihood. Only computed when robust=1.
D2linv

inverse of the derivative of the score function.
score

value of score for final estimates.
test.procProp

observed score process for proportional part of model.
var.score

variance of score process (optional variation estimator for beta.fixed=1 and robust estimator otherwise).
pval.Prop

p-value based on resampling.
sim.supProp

re-sampled absolute supremum values.
sim.test.procProp

list of 50 random realizations of test-processes for proportionality under the model based on resampling.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

library(timereg)
data(sTRACE)
# Fits Cox model 
out<-cox.aalen(Surv(time,status==9)~prop(age)+prop(sex)+
prop(vf)+prop(chf)+prop(diabetes),data=sTRACE)

# makes Lin, Wei, Ying test for proportionality
summary(out)
par(mfrow=c(2,3))
plot(out,score=1) 

# Fits stratified Cox model 
out<-cox.aalen(Surv(time,status==9)~-1+factor(vf)+ prop(age)+prop(sex)+
	       prop(chf)+prop(diabetes),data=sTRACE,max.time=7,n.sim=100)
summary(out)
par(mfrow=c(1,2)); plot(out); 
# Same model, but needs to invert the entire marix for the aalen part: X(t) 
out<-cox.aalen(Surv(time,status==9)~factor(vf)+ prop(age)+prop(sex)+
	       prop(chf)+prop(diabetes),data=sTRACE,max.time=7,n.sim=100)
summary(out)
par(mfrow=c(1,2)); plot(out); 


# Fits Cox-Aalen model 
out<-cox.aalen(Surv(time,status==9)~prop(age)+prop(sex)+
               vf+chf+prop(diabetes),data=sTRACE,max.time=7,n.sim=100)
summary(out)
par(mfrow=c(2,3))
plot(out)

Missing data IPW Cox

Description

Fits an Cox-Aalen survival model with missing data, with glm specification of probability of missingness.

Usage

cox.ipw(
  survformula,
  glmformula,
  d = parent.frame(),
  max.clust = NULL,
  ipw.se = FALSE,
  tie.seed = 100
)

Arguments

survformula

a formula object with the response on the left of a '~' operator, and the independent terms on the right as regressors. The response must be a survival object as returned by the ‘Surv’ function.

Adds the prop() wrapper internally for using cox.aalen function for fitting Cox model.
glmformula

formula for "being" observed, that is not missing.
d

data frame.
max.clust

number of clusters in iid approximation. Default is all.
ipw.se

if TRUE computes standard errors based on iid decompositon of cox and glm model, thus should be asymptotically correct.
tie.seed

if there are ties these are broken, and to get same break the seed must be the same. Recommend to break them prior to entering the program.

Details

Taylor expansion of Cox's partial likelihood in direction of glm parameters using num-deriv and iid expansion of Cox and glm paramters (lava).

Value

returns an object of type "cox.aalen". With the following arguments:

iid

iid decomposition.
coef

missing data estiamtes for weighted cox.

var

robust pointwise variances estimates.

se

robust pointwise variances estimates.

se.naive

estimate of parametric components of model.

ties

list of ties and times with random noise to break ties.
cox

output from weighted cox model.

Author(s)

Thomas Scheike

References

Paik et al.

Examples

### fit <- cox.ipw(Surv(time,status)~X+Z,obs~Z+X+time+status,data=d,ipw.se=TRUE)
### summary(fit)

CSL liver chirrosis data

Description

Survival status for the liver chirrosis patients of Schlichting et al.

Format

This data frame contains the following columns:

id

a numeric vector. Id of subject.

time

a numeric vector. Time of measurement.

prot

a numeric vector. Prothrombin level at measurement time.

dc

a numeric vector code. 0: censored observation, 1: died at eventT.

eventT

a numeric vector. Time of event (death).

treat

a numeric vector code. 0: active treatment of prednisone, 1: placebo treatment.

sex

a numeric vector code. 0: female, 1: male.

age

a numeric vector. Age of subject at inclusion time subtracted 60.

prot.base

a numeric vector. Prothrombin base level before entering the study.

prot.prev

a numeric vector. Level of prothrombin at previous measurement time.

lt

a numeric vector. Gives the starting time for the time-intervals.

rt

a numeric vector. Gives the stopping time for the time-intervals.

Source

P.K. Andersen

References

Schlichting, P., Christensen, E., Andersen, P., Fauerholds, L., Juhl, E., Poulsen, H. and Tygstrup, N. (1983), The Copenhagen Study Group for Liver Diseases, Hepatology 3, 889–895

Examples

data(csl)
names(csl)

Model validation based on cumulative residuals

Description

Computes cumulative residuals and approximative p-values based on resampling techniques.

Usage

cum.residuals(
  object,
  data = parent.frame(),
  modelmatrix = 0,
  cum.resid = 1,
  n.sim = 500,
  weighted.test = 0,
  max.point.func = 50,
  weights = NULL
)

Arguments

object

an object of class 'aalen', 'timecox', 'cox.aalen' where the residuals are returned ('residuals=1')
data

data frame based on which residuals are computed.
modelmatrix

specifies a grouping of the data that is used for cumulating residuals. Must have same size as data and be ordered in the same way.
cum.resid

to compute residuals versus each of the continuous covariates in the model.
n.sim

number of simulations in resampling.
weighted.test

to compute a variance weighted version of the test-processes used for testing constant effects of covariates.
max.point.func

limits the amount of computations, only considers a max of 50 points on the covariate scales.
weights

weights for sum of martingale residuals, now for cum.resid=1.

Value

returns an object of type "cum.residuals" with the following arguments:

cum

cumulative residuals versus time for the groups specified by modelmatrix.

var.cum

the martingale based pointwise variance estimates.
robvar.cum

robust pointwise variances estimates of cumulatives.
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value covariate effects based on supremum test.
sim.testBeq0

resampled supremum value.
conf.band

resampling based constant to construct robust 95% uniform confidence bands for cumulative residuals.
obs.test

absolute value of supremum of observed test-process.
pval.test

p-value for supremum test statistic.
sim.test

resampled absolute value of supremum cumulative residuals.
proc.cumz

observed cumulative residuals versus all continuous covariates of model.
sim.test.proccumz

list of 50 random realizations of test-processes under model for all continuous covariates.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

data(sTRACE)
# Fits Aalen model and returns residuals
fit<-aalen(Surv(time,status==9)~age+sex+diabetes+chf+vf,
           data=sTRACE,max.time=7,n.sim=0,residuals=1)

# constructs and simulates cumulative residuals versus age groups
fit.mg<-cum.residuals(fit,data=sTRACE,n.sim=100,
modelmatrix=model.matrix(~-1+factor(cut(age,4)),sTRACE))

par(mfrow=c(1,4))
# cumulative residuals with confidence intervals
plot(fit.mg);
# cumulative residuals versus processes under model
plot(fit.mg,score=1); 
summary(fit.mg)

# cumulative residuals vs. covariates Lin, Wei, Ying style 
fit.mg<-cum.residuals(fit,data=sTRACE,cum.resid=1,n.sim=100)

par(mfrow=c(2,4))
plot(fit.mg,score=2)
summary(fit.mg)

The Diabetic Retinopathy Data

Description

The data was colleceted to test a laser treatment for delaying blindness in patients with dibetic retinopathy. The subset of 197 patiens given in Huster et al. (1989) is used.

Format

This data frame contains the following columns:

id

a numeric vector. Patient code.

agedx

a numeric vector. Age of patient at diagnosis.

time

a numeric vector. Survival time: time to blindness or censoring.

status

a numeric vector code. Survival status. 1: blindness, 0: censored.

trteye

a numeric vector code. Random eye selected for treatment. 1: left eye 2: right eye.

treat

a numeric vector. 1: treatment 0: untreated.

adult

a numeric vector code. 1: younger than 20, 2: older than 20.

Source

Huster W.J. and Brookmeyer, R. and Self. S. (1989) MOdelling paired survival data with covariates, Biometrics 45, 145-56.

Examples

data(diabetes)
names(diabetes)

Fit time-varying regression model

Description

Fits time-varying regression model with partly parametric components. Time-dependent variables for longitudinal data. The model assumes that the mean of the observed responses given covariates is a linear time-varying regression model :

Usage

dynreg(
  formula = formula(data),
  data = parent.frame(),
  aalenmod,
  bandwidth = 0.5,
  id = NULL,
  bhat = NULL,
  start.time = 0,
  max.time = NULL,
  n.sim = 500,
  meansub = 1,
  weighted.test = 0,
  resample = 0
)

Arguments

formula

a formula object with the response on the left of a '~' operator, and the independent terms on the right as regressors.
data

a data.frame with the variables.
aalenmod

Aalen model for measurement times. Specified as a survival model (see aalen function).
bandwidth

bandwidth for local iterations. Default is 50% of the range of the considered observation period.
id

For timevarying covariates the variable must associate each record with the id of a subject.
bhat

initial value for estimates. If NULL local linear estimate is computed.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. Default is max of data.
n.sim

number of simulations in resampling.
meansub

if '1' then the mean of the responses is subtracted before the estimation is carried out.
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
resample

returns resample processes.

Details

E(Zij∣Xij(t))=βT(t)Xij1(t)+γTXij2(t)E( Z_{ij} | X_{ij}(t) ) = \beta^T(t) X_{ij}^1(t) + \gamma^T X_{ij}^2(t)

where ZijZ_{ij} is the j'th measurement at time t for the i'th subject with covariates Xij1X_{ij}^1 and Xij2X_{ij}^2. Resampling is used for computing p-values for tests of timevarying effects. The data for a subject is presented as multiple rows or 'observations', each of which applies to an interval of observation (start, stop]. For counting process data with the )start,stop] notation is used the 'id' variable is needed to identify the records for each subject. The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Value

returns an object of type "dynreg". With the following arguments:

cum

the cumulative regression coefficients. This is the efficient estimator based on an initial smoother obtained by local linear regression :

B^(t)=∫0tβ~(s)ds+\hat B(t) = \int_0^t \tilde \beta(s) ds+ \hspace{4 cm}

∫0tX−(Diag(z)−Diag(XT(s)β~(s)))dp(ds×dz),\int_0^t X^{-} (Diag(z) -Diag( X^T(s) \tilde \beta(s)) ) dp(ds \times dz),

where β~(t)\tilde \beta(t) is an initial estimate either provided or computed by local linear regression. To plot this estimate use type="eff.smooth" in the plot() command.

var.cum

the martingale based pointwise variance estimates.
robvar.cum

robust pointwise variances estimates.
gamma

estimate of semi-parametric components of model.
var.gamma

variance for gamma.
robvar.gamma

robust variance for gamma.
cum0

simple estimate of cumulative regression coefficients that does not use use an initial smoothing based estimate

B^0(t)=∫0tX−Diag(z)dp(ds×dz).\hat B_0(t) = \int_0^t X^{-} Diag(z) dp(ds \times dz).

To plot this estimate use type="0.mpp" in the plot() command.

var.cum0

the martingale based pointwise variance estimates of cum0.
cum.ms

estimate of cumulative regression coefficients based on initial smoother (but robust to this estimator).

B^ms(t)=∫0tX−(Diag(z)−f(s))dp(ds×dz),\hat B_{ms}(t) = \int_0^t X^{-} (Diag(z)-f(s)) dp(ds \times dz),

where ff is chosen as the matrix

f(s)=Diag(XT(s)β~(s))(I−Xα(s)Xα−(s)),f(s) = Diag( X^T(s) \tilde \beta(s)) ( I - X_\alpha(s) X_\alpha^-(s) ),

where XαX_{\alpha} is the design for the sampling intensities. This is also an efficient estimator when the initial estimator is consistent for β(t)\beta(t) and then asymptotically equivalent to cum, but small sample properties appear inferior. Its variance is estimated by var.cum. To plot this estimate use type="ms.mpp" in the plot() command.

cum.ly

estimator where local averages are subtracted. Special case of cum.ms. To plot this estimate use type="ly.mpp" in plot.

var.cum.ly

the martingale based pointwise variance estimates.

gamma0

estimate of parametric component of model.

var.gamma0

estimate of variance of parametric component of model.

gamma.ly

estimate of parametric components of model.

var.gamma.ly

estimate of variance of parametric component of model.

gamma.ms

estimate of variance of parametric component of model.

var.gamma.ms

estimate of variance of parametric component of model.
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.
test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect.
sim.test.procBeqC

list of 50 random realizations of test-processes under null based on resampling.
covariance

covariances for nonparametric terms of model.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

## this runs slowly and is therfore donttest
data(csl)
indi.m<-rep(1,length(csl$lt)) 

# Fits time-varying regression model 
out<-dynreg(prot~treat+prot.prev+sex+age,data=csl,
Surv(lt,rt,indi.m)~+1,start.time=0,max.time=2,id=csl$id,
n.sim=100,bandwidth=0.7,meansub=0)
summary(out)
par(mfrow=c(2,3))
plot(out)

# Fits time-varying semi-parametric regression model.
outS<-dynreg(prot~treat+const(prot.prev)+const(sex)+const(age),data=csl,
Surv(lt,rt,indi.m)~+1,start.time=0,max.time=2,id=csl$id,
n.sim=100,bandwidth=0.7,meansub=0)
summary(outS)

Event history object

Description

Constructur for Event History objects

Usage

Event(time, time2 = TRUE, cause = NULL, cens.code = 0, ...)

Arguments

time

Time
time2

Time 2
cause

Cause
cens.code

Censoring code (default 0)
...

Additional arguments

Details

... content for details

Value

Object of class Event (a matrix)

Author(s)

Klaus K. Holst and Thomas Scheike

Examples

t1 <- 1:10
	t2 <- t1+runif(10)
	ca <- rbinom(10,2,0.4)
	(x <- Event(t1,t2,ca))

EventSplit (SurvSplit).

Description

contstructs start stop formulation of event time data after a variable in the data.set. Similar to SurvSplit of the survival package but can also split after random time given in data frame.

Usage

event.split(
  data,
  time = "time",
  status = "status",
  cuts = "cuts",
  name.id = "id",
  name.start = "start",
  cens.code = 0,
  order.id = TRUE,
  time.group = FALSE
)

Arguments

data

data to be split
time

time variable.
status

status variable.
cuts

cuts variable or numeric cut (only one value)
name.id

name of id variable.
name.start

name of start variable in data, start can also be numeric "0"
cens.code

code for the censoring.
order.id

order data after id and start.
time.group

make variable "before"."cut" that keeps track of wether start,stop is before (1) or after cut (0).

Author(s)

Thomas Scheike

Examples

set.seed(1)
d <- data.frame(event=round(5*runif(5),2),start=1:5,time=2*1:5,
		status=rbinom(5,1,0.5),x=1:5)
d

d0 <- event.split(d,cuts="event",name.start=0)
d0

dd <- event.split(d,cuts="event")
dd
ddd <- event.split(dd,cuts=3.5)
ddd
event.split(ddd,cuts=5.5)

### successive cutting for many values 
dd <- d
for  (cuts in seq(2,3,by=0.3)) dd <- event.split(dd,cuts=cuts)
dd

###########################################################################
### same but for situation with multiple events along the time-axis
###########################################################################
d <- data.frame(event1=1:5+runif(5)*0.5,start=1:5,time=2*1:5,
		status=rbinom(5,1,0.5),x=1:5,start0=0)
d$event2 <- d$event1+0.2
d$event2[4:5] <- NA 
d

d0 <- event.split(d,cuts="event1",name.start="start",time="time",status="status")
d0
###
d00 <- event.split(d0,cuts="event2",name.start="start",time="time",status="status")
d00

Fit Generalized Semiparametric Proportional 0dds Model

Description

Fits a semiparametric proportional odds model:

logit(1−SX,Z(t))=log(XTA(t))+βTZlogit(1-S_{X,Z}(t)) = log(X^T A(t)) + \beta^T Z

where A(t) is increasing but otherwise unspecified. Model is fitted by maximising the modified partial likelihood. A goodness-of-fit test by considering the score functions is also computed by resampling methods.

Usage

Gprop.odds(
  formula = formula(data),
  data = parent.frame(),
  beta = 0,
  Nit = 50,
  detail = 0,
  start.time = 0,
  max.time = NULL,
  id = NULL,
  n.sim = 500,
  weighted.test = 0,
  sym = 0,
  mle.start = 0
)

Arguments

formula

a formula object, with the response on the left of a '~' operator, and the terms on the right. The response must be a survival object as returned by the ‘Surv’ function.
data

a data.frame with the variables.
beta

starting value for relative risk estimates
Nit

number of iterations for Newton-Raphson algorithm.
detail

if 0 no details is printed during iterations, if 1 details are given.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. This is very useful to obtain stable estimates, especially for the baseline. Default is max of data.
id

For timevarying covariates the variable must associate each record with the id of a subject.
n.sim

number of simulations in resampling.
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
sym

to use symmetrized second derivative in the case of the estimating equation approach (profile=0). This may improve the numerical performance.
mle.start

starting values for relative risk parameters.

Details

An alternative way of writing the model :

SX,Z(t))=exp⁡(−βTZ)(XTA(t))+exp⁡(−βTZ)S_{X,Z}(t)) = \frac{ \exp( - \beta^T Z )}{ (X^T A(t)) + \exp( - \beta^T Z) }

such that β\beta is the log-odds-ratio of dying before time t, and A(t)A(t) is the odds-ratio.

The modelling formula uses the standard survival modelling given in the survival package.

The data for a subject is presented as multiple rows or "observations", each of which applies to an interval of observation (start, stop]. The program essentially assumes no ties, and if such are present a little random noise is added to break the ties.

Value

returns an object of type 'cox.aalen'. With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

the martingale based pointwise variance estimates.

robvar.cum

robust pointwise variances estimates.

gamma

estimate of proportional odds parameters of model.
var.gamma

variance for gamma.

robvar.gamma

robust variance for gamma.

residuals

list with residuals. Estimated martingale increments (dM) and corresponding time vector (time).
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.

test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect over time.
loglike

modified partial likelihood, pseudo profile likelihood for regression parameters.
D2linv

inverse of the derivative of the score function.
score

value of score for final estimates.
test.procProp

observed score process for proportional odds regression effects.
pval.Prop

p-value based on resampling.
sim.supProp

re-sampled supremum values.
sim.test.procProp

list of 50 random realizations of test-processes for constant proportional odds under the model based on resampling.

Author(s)

Thomas Scheike

References

Scheike, A flexible semiparametric transformation model for survival data, Lifetime Data Anal. (to appear).

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

data(sTRACE)

### runs slowly and is therefore donttest 
data(sTRACE)
# Fits Proportional odds model with stratified baseline
age.c<-scale(sTRACE$age,scale=FALSE); 
## out<-Gprop.odds(Surv(time,status==9)~-1+factor(diabetes)+prop(age.c)+prop(chf)+
##                 prop(sex)+prop(vf),data=sTRACE,max.time=7,n.sim=50)
## summary(out) 
## par(mfrow=c(2,3))
## plot(out,sim.ci=2); plot(out,score=1)

Fits Krylow based PLS for additive hazards model

Description

Fits the PLS estimator for the additive risk model based on the least squares fitting criterion

Usage

krylow.pls(D, d, dim = 1)

Arguments

D

defined above
d

defined above
dim

number of pls dimensions

Details

L(β,D,d)=βTDβ−2βTdL(\beta,D,d) = \beta^T D \beta - 2 \beta^T d

where D=∫ZHZdtD=\int Z H Z dt and d=∫ZHdNd=\int Z H dN.

Value

returns a list with the following arguments:

beta

PLS regression coefficients

Author(s)

Thomas Scheike

References

Martinussen and Scheike, The Aalen additive hazards model with high-dimensional regressors, submitted.

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

## makes data for pbc complete case
data(mypbc)
pbc<-mypbc
pbc$time<-pbc$time+runif(418)*0.1; pbc$time<-pbc$time/365
pbc<-subset(pbc,complete.cases(pbc));
covs<-as.matrix(pbc[,-c(1:3,6)])
covs<-cbind(covs[,c(1:6,16)],log(covs[,7:15]))

## computes the matrices needed for the least squares 
## criterion 
out<-aalen(Surv(time,status>=1)~const(covs),pbc,robust=0,n.sim=0)
S=out$intZHZ; s=out$intZHdN;

out<-krylow.pls(S,s,dim=2)

Melanoma data and Danish population mortality by age and sex

Description

Melanoma data with background mortality of Danish population.

Format

This data frame contains the following columns:

id

a numeric vector. Gives patient id.

sex

a numeric vector. Gives sex of patient.

start

a numeric vector. Gives the starting time for the time-interval for which the covariate rate is representative.

stop

a numeric vector. Gives the stopping time for the time-interval for which the covariate rate is representative.

status

a numeric vector code. Survival status. 1: dead from melanoma, 0: alive or dead from other cause.

age

a numeric vector. Gives the age of the patient at removal of tumor.

rate

a numeric vector. Gives the population mortality for the given sex and age. Based on Table A.2 in Andersen et al. (1993).

Source

Andersen, P.K., Borgan O, Gill R.D., Keiding N. (1993), Statistical Models Based on Counting Processes, Springer-Verlag.

Examples

data(mela.pop)
names(mela.pop)

The Melanoma Survival Data

Description

The melanoma data frame has 205 rows and 7 columns. It contains data relating to survival of patients after operation for malignant melanoma collected at Odense University Hospital by K.T. Drzewiecki.

Format

This data frame contains the following columns:

no

a numeric vector. Patient code.

status

a numeric vector code. Survival status. 1: dead from melanoma, 2: alive, 3: dead from other cause.

days

a numeric vector. Survival time.

ulc

a numeric vector code. Ulceration, 1: present, 0: absent.

thick

a numeric vector. Tumour thickness (1/100 mm).

sex

a numeric vector code. 0: female, 1: male.

Source

Andersen, P.K., Borgan O, Gill R.D., Keiding N. (1993), Statistical Models Based on Counting Processes, Springer-Verlag.

Drzewiecki, K.T., Ladefoged, C., and Christensen, H.E. (1980), Biopsy and prognosis for cutaneous malignant melanoma in clinical stage I. Scand. J. Plast. Reconstru. Surg. 14, 141-144.

Examples

data(melanoma)
names(melanoma)

my version of the PBC data of the survival package

Description

my version of the PBC data of the survival package

Source

survival package

Make predictions of predict functions in rows mononotone

Description

Make predictions of predict functions in rows mononotone using the pool-adjacent-violators-algorithm

Usage

pava.pred(pred, increasing = TRUE)

Arguments

pred

predictions, either vector or rows of predictions.
increasing

increasing or decreasing.

Value

mononotone predictions.

Author(s)

Thomas Scheike

Examples

data(bmt); 

## competing risks 
add<-comp.risk(Event(time,cause)~platelet+age+tcell,data=bmt,cause=1)
ndata<-data.frame(platelet=c(1,0,0),age=c(0,1,0),tcell=c(0,0,1))
out<-predict(add,newdata=ndata,uniform=0)

par(mfrow=c(1,1))
head(out$P1)
matplot(out$time,t(out$P1),type="s")

###P1m <- t(apply(out$P1,1,pava))
P1monotone <- pava.pred(out$P1)
head(P1monotone)
matlines(out$time,t(P1monotone),type="s")

Fits Proportional excess hazards model with fixed offsets

Description

Fits proportional excess hazards model. The Sasieni proportional excess risk model.

Usage

pe.sasieni(
  formula = formula(data),
  data = parent.frame(),
  id = NULL,
  start.time = 0,
  max.time = NULL,
  offsets = 0,
  Nit = 50,
  detail = 0,
  n.sim = 500
)

Arguments

formula

a formula object, with the response on the left of a ‘~’ operator, and the terms on the right. The response must be a survival object as returned by the ‘Surv’ function.
data

a data.frame with the variables.
id

gives the number of individuals.
start.time

starting time for considered time-period.
max.time

stopping considered time-period if different from 0. Estimates thus computed from [0,max.time] if max.time>0. Default is max of data.
offsets

fixed offsets giving the mortality.
Nit

number of itterations.
detail

if detail is one, prints iteration details.
n.sim

number of simulations, 0 for no simulations.

Details

The models are written using the survival modelling given in the survival package.

The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Value

Returns an object of type "pe.sasieni". With the following arguments:

cum

baseline of Cox model excess risk.
var.cum

pointwise variance estimates for estimated cumulatives.
gamma

estimate of relative risk terms of model.
var.gamma

variance estimates for gamma.
Ut

score process for Cox part of model.
D2linv

The inverse of the second derivative.
score

final score
test.Prop

re-sampled absolute supremum values.
pval.Prop

p-value based on resampling.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer Verlag (2006).

Sasieni, P.D., Proportional excess hazards, Biometrika (1996), 127–41.

Cortese, G. and Scheike, T.H., Dynamic regression hazards models for relative survival (2007), submitted.

Examples

data(mela.pop)
out<-pe.sasieni(Surv(start,stop,status==1)~age+sex,mela.pop,
id=1:205,Nit=10,max.time=7,offsets=mela.pop$rate,detail=0,n.sim=100)
summary(out)

ul<-out$cum[,2]+1.96*out$var.cum[,2]^.5
ll<-out$cum[,2]-1.96*out$var.cum[,2]^.5
plot(out$cum,type="s",ylim=range(ul,ll))
lines(out$cum[,1],ul,type="s"); lines(out$cum[,1],ll,type="s")
# see also prop.excess function

Plots estimates and test-processes

Description

This function plots the non-parametric cumulative estimates for the additive risk model or the test-processes for the hypothesis of time-varying effects with re-sampled processes under the null.

Usage

## S3 method for class 'aalen'
plot(
  x,
  pointwise.ci = 1,
  hw.ci = 0,
  sim.ci = 0,
  robust.ci = 0,
  col = NULL,
  specific.comps = FALSE,
  level = 0.05,
  start.time = 0,
  stop.time = 0,
  add.to.plot = FALSE,
  mains = TRUE,
  xlab = "Time",
  ylab = "Cumulative coefficients",
  score = FALSE,
  ...
)

Arguments

x

the output from the "aalen" function.
pointwise.ci

if >1 pointwise confidence intervals are plotted with lty=pointwise.ci
hw.ci

if >1 Hall-Wellner confidence bands are plotted with lty=hw.ci. Only 0.95 % bands can be constructed.
sim.ci

if >1 simulation based confidence bands are plotted with lty=sim.ci. These confidence bands are robust to non-martingale behaviour.
robust.ci

robust standard errors are used to estimate standard error of estimate, otherwise martingale based standard errors are used.
col

specifice colors of different components of plot, in order: c(estimate,pointwise.ci,robust.ci,hw.ci,sim.ci) so for example, when we ask to get pointwise.ci, hw.ci and sim.ci we would say c(1,2,3,4) to use colors as specified.
specific.comps

all components of the model is plotted by default, but a list of components may be specified, for example first and third "c(1,3)".
level

gives the significance level.
start.time

start of observation period where estimates are plotted.
stop.time

end of period where estimates are plotted. Estimates thus plotted from [start.time, max.time].
add.to.plot

to add to an already existing plot.
mains

add names of covariates as titles to plots.
xlab

label for x-axis.
ylab

label for y-axis.
score

to plot test processes for test of time-varying effects along with 50 random realization under the null-hypothesis.
...

unused arguments - for S3 compatibility

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression models for Survival Data, Springer (2006).

Examples

# see help(aalen) 
data(sTRACE)
out<-aalen(Surv(time,status==9)~chf+vf,sTRACE,max.time=7,n.sim=100)
par(mfrow=c(2,2))
plot(out,pointwise.ci=1,hw.ci=1,sim.ci=1,col=c(1,2,3,4))
par(mfrow=c(2,2))
plot(out,pointwise.ci=0,robust.ci=1,hw.ci=1,sim.ci=1,col=c(1,2,3,4))

Plots cumulative residuals

Description

This function plots the output from the cumulative residuals function "cum.residuals". The cumulative residuals are compared with the performance of similar processes under the model.

Usage

## S3 method for class 'cum.residuals'
plot(
  x,
  pointwise.ci = 1,
  hw.ci = 0,
  sim.ci = 0,
  robust = 1,
  specific.comps = FALSE,
  level = 0.05,
  start.time = 0,
  stop.time = 0,
  add.to.plot = FALSE,
  mains = TRUE,
  main = NULL,
  xlab = NULL,
  ylab = "Cumulative MG-residuals",
  ylim = NULL,
  score = 0,
  conf.band = FALSE,
  ...
)

Arguments

x

the output from the "cum.residuals" function.
pointwise.ci

if >1 pointwise confidence intervals are plotted with lty=pointwise.ci
hw.ci

if >1 Hall-Wellner confidence bands are plotted with lty=hw.ci. Only 95% bands can be constructed.
sim.ci

if >1 simulation based confidence bands are plotted with lty=sim.ci. These confidence bands are robust to non-martingale behaviour.
robust

if "1" robust standard errors are used to estimate standard error of estimate, otherwise martingale based estimate are used.
specific.comps

all components of the model is plotted by default, but a list of components may be specified, for example first and third "c(1,3)".
level

gives the significance level. Default is 0.05.
start.time

start of observation period where estimates are plotted. Default is 0.
stop.time

end of period where estimates are plotted. Estimates thus plotted from [start.time, max.time].
add.to.plot

to add to an already existing plot. Default is "FALSE".
mains

add names of covariates as titles to plots.
main

vector of names for titles in plots.
xlab

label for x-axis. NULL is default which leads to "Time" or "". Can also give a character vector.
ylab

label for y-axis. Default is "Cumulative MG-residuals".
ylim

limits for y-axis.
score

if '0' plots related to modelmatrix are specified, thus resulting in grouped residuals, if '1' plots for modelmatrix but with random realizations under model, if '2' plots residuals versus continuous covariates of model with random realizations under the model.
conf.band

makes simulation based confidence bands for the test processes under the 0 based on variance of these processes limits for y-axis. These will give additional information of whether the observed cumulative residuals are extreme or not when based on a variance weighted test.
...

unused arguments - for S3 compatibility

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

# see cum.residuals for examples

Plots estimates and test-processes

Description

This function plots the non-parametric cumulative estimates for the additive risk model or the test-processes for the hypothesis of constant effects with re-sampled processes under the null.

Usage

## S3 method for class 'dynreg'
plot(
  x,
  type = "eff.smooth",
  pointwise.ci = 1,
  hw.ci = 0,
  sim.ci = 0,
  robust = 0,
  specific.comps = FALSE,
  level = 0.05,
  start.time = 0,
  stop.time = 0,
  add.to.plot = FALSE,
  mains = TRUE,
  xlab = "Time",
  ylab = "Cumulative coefficients",
  score = FALSE,
  ...
)

Arguments

x

the output from the "dynreg" function.
type

the estimator plotted. Choices "eff.smooth", "ms.mpp", "0.mpp" and "ly.mpp". See the dynreg function for more on this.
pointwise.ci

if >1 pointwise confidence intervals are plotted with lty=pointwise.ci
hw.ci

if >1 Hall-Wellner confidence bands are plotted with lty=hw.ci. Only 0.95 % bands can be constructed.
sim.ci

if >1 simulation based confidence bands are plotted with lty=sim.ci. These confidence bands are robust to non-martingale behaviour.
robust

robust standard errors are used to estimate standard error of estimate, otherwise martingale based estimate are used.
specific.comps

all components of the model is plotted by default, but a list of components may be specified, for example first and third "c(1,3)".
level

gives the significance level.
start.time

start of observation period where estimates are plotted.
stop.time

end of period where estimates are plotted. Estimates thus plotted from [start.time, max.time].
add.to.plot

to add to an already existing plot.
mains

add names of covariates as titles to plots.
xlab

label for x-axis.
ylab

label for y-axis.
score

to plot test processes for test of time-varying effects along with 50 random realization under the null-hypothesis.
...

unused arguments - for S3 compatibility

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

### runs slowly and therefore donttest 
data(csl)
indi.m<-rep(1,length(csl$lt))

# Fits time-varying regression model
out<-dynreg(prot~treat+prot.prev+sex+age,csl,
Surv(lt,rt,indi.m)~+1,start.time=0,max.time=3,id=csl$id,
n.sim=100,bandwidth=0.7,meansub=0)

par(mfrow=c(2,3))
# plots estimates 
plot(out)
# plots tests-processes for time-varying effects 
plot(out,score=TRUE)

Predictions for Survival and Competings Risks Regression for timereg

Description

Make predictions based on the survival models (Aalen and Cox-Aalen) and the competing risks models for the cumulative incidence function (comp.risk). Computes confidence intervals and confidence bands based on resampling.

Usage

## S3 method for class 'timereg'
predict(
  object,
  newdata = NULL,
  X = NULL,
  times = NULL,
  Z = NULL,
  n.sim = 500,
  uniform = TRUE,
  se = TRUE,
  alpha = 0.05,
  resample.iid = 0,
  ...
)

Arguments

object

an object belonging to one of the following classes: comprisk, aalen or cox.aalen
newdata

specifies the data at which the predictions are wanted.
X

alternative to newdata, specifies the nonparametric components for predictions.
times

times in which predictions are computed, default is all time-points for baseline
Z

alternative to newdata, specifies the parametric components of the model for predictions.
n.sim

number of simulations in resampling.
uniform

computes resampling based uniform confidence bands.
se

computes pointwise standard errors
alpha

specificies the significance levelwhich cause we consider.
resample.iid

set to 1 to return iid decomposition of estimates, 3-dim matrix (predictions x times x subjects)
...

unused arguments - for S3 compatability

Value

time

vector of time points where the predictions are computed.
unif.band

resampling based constant to construct 95% uniform confidence bands.
model

specifies what model that was fitted.
alpha

specifies the significance level for the confidence intervals. This relates directly to the constant given in unif.band.
newdata

specifies the newdata given in the call.
RR

gives relative risk terms for Cox-type models.
call

gives call for predict funtion.
initial.call

gives call for underlying object used for predictions.
P1

gives cumulative inicidence predictions for competing risks models. Predictions given in matrix form with different subjects in different rows.
S0

gives survival predictions for survival models. Predictions given in matrix form with different subjects in different rows.
se.P1

pointwise standard errors for predictions of P1.
se.S0

pointwise standard errors for predictions of S0.

Author(s)

Thomas Scheike, Jeremy Silver

References

Scheike, Zhang and Gerds (2008), Predicting cumulative incidence probability by direct binomial regression, Biometrika, 95, 205-220.

Scheike and Zhang (2007), Flexible competing risks regression modelling and goodness of fit, LIDA, 14, 464-483 .

Martinussen and Scheike (2006), Dynamic regression models for survival data, Springer.

Examples

data(bmt); 

## competing risks 
add<-comp.risk(Event(time,cause)~platelet+age+tcell,data=bmt,cause=1)

ndata<-data.frame(platelet=c(1,0,0),age=c(0,1,0),tcell=c(0,0,1))
out<-predict(add,newdata=ndata,uniform=1,n.sim=1000)
par(mfrow=c(2,2))
plot(out,multiple=0,uniform=1,col=1:3,lty=1,se=1)
# see comp.risk for further examples. 

add<-comp.risk(Event(time,cause)~factor(tcell),data=bmt,cause=1)
summary(add)
out<-predict(add,newdata=ndata,uniform=1,n.sim=1000)
plot(out,multiple=1,uniform=1,col=1:3,lty=1,se=1)

add<-prop.odds.subdist(Event(time,cause)~factor(tcell),
	       data=bmt,cause=1)
out <- predict(add,X=1,Z=1)
plot(out,multiple=1,uniform=1,col=1:3,lty=1,se=1)


## SURVIVAL predictions aalen function
data(sTRACE)
out<-aalen(Surv(time,status==9)~sex+ diabetes+chf+vf,
data=sTRACE,max.time=7,n.sim=0,resample.iid=1)

pout<-predict(out,X=rbind(c(1,0,0,0,0),rep(1,5)))
head(pout$S0[,1:5]); head(pout$se.S0[,1:5])
par(mfrow=c(2,2))
plot(pout,multiple=1,se=0,uniform=0,col=1:2,lty=1:2)
plot(pout,multiple=0,se=1,uniform=1,col=1:2)

out<-aalen(Surv(time,status==9)~const(age)+const(sex)+
const(diabetes)+chf+vf,
data=sTRACE,max.time=7,n.sim=0,resample.iid=1)

pout<-predict(out,X=rbind(c(1,0,0),c(1,1,0)),
Z=rbind(c(55,0,1),c(60,1,1)))
head(pout$S0[,1:5]); head(pout$se.S0[,1:5])
par(mfrow=c(2,2))
plot(pout,multiple=1,se=0,uniform=0,col=1:2,lty=1:2)
plot(pout,multiple=0,se=1,uniform=1,col=1:2)

pout<-predict(out,uniform=0,se=0,newdata=sTRACE[1:10,]) 
plot(pout,multiple=1,se=0,uniform=0)

#### cox.aalen
out<-cox.aalen(Surv(time,status==9)~prop(age)+prop(sex)+
prop(diabetes)+chf+vf,
data=sTRACE,max.time=7,n.sim=0,resample.iid=1)

pout<-predict(out,X=rbind(c(1,0,0),c(1,1,0)),Z=rbind(c(55,0,1),c(60,1,1)))
head(pout$S0[,1:5]); head(pout$se.S0[,1:5])
par(mfrow=c(2,2))
plot(pout,multiple=1,se=0,uniform=0,col=1:2,lty=1:2)
plot(pout,multiple=0,se=1,uniform=1,col=1:2)

pout<-predict(out,uniform=0,se=0,newdata=sTRACE[1:10,]) 
plot(pout,multiple=1,se=0,uniform=0)

#### prop.odds model 
add<-prop.odds(Event(time,cause!=0)~factor(tcell),data=bmt)
out <- predict(add,X=1,Z=0)
plot(out,multiple=1,uniform=1,col=1:3,lty=1,se=1)

Set up weights for delayed-entry competing risks data for comp.risk function

Description

Computes the weights of Geskus (2011) modified to the setting of the comp.risk function. The returned weights are 1/(H(Ti)∗Gc(min(Ti,tau)))1/(H(T_i)*G_c(min(T_i,tau))) and tau is the max of the times argument, here HH is the estimator of the truncation distribution and GcG_c is the right censoring distribution.

Usage

prep.comp.risk(
  data,
  times = NULL,
  entrytime = NULL,
  time = "time",
  cause = "cause",
  cname = "cweight",
  tname = "tweight",
  strata = NULL,
  nocens.out = TRUE,
  cens.formula = NULL,
  cens.code = 0,
  prec.factor = 100,
  trunc.mintau = FALSE
)

Arguments

data

data frame for comp.risk.
times

times for estimating equations.
entrytime

name of delayed entry variable, if not given computes right-censoring case.
time

name of survival time variable.
cause

name of cause indicator
cname

name of censoring weight.
tname

name of truncation weight.
strata

strata variable to obtain stratified weights.
nocens.out

returns only uncensored part of data-frame
cens.formula

censoring model formula for Cox models for the truncation and censoring model.
cens.code

code for censoring among causes.
prec.factor

precision factor, for ties between censoring/even times, truncation times/event times
trunc.mintau

specicies wether the truncation distribution is evaluated in death times or death times minimum max(times), FALSE makes the estimator equivalent to Kaplan-Meier (in the no covariate case).

Value

Returns an object. With the following arguments:

dataw

a data.frame with weights.

The function wants to make two new variables "weights" and "cw" so if these already are in the data frame it tries to add an "_" in the names.

Author(s)

Thomas Scheike

References

Geskus (2011), Cause-Specific Cumulative Incidence Estimation and the Fine and Gray Model Under Both Left Truncation and Right Censoring, Biometrics (2011), pp 39-49.

Shen (2011), Proportional subdistribution hazards regression for left-truncated competing risks data, Journal of Nonparametric Statistics (2011), 23, 885-895

Examples

data(bmt)
nn <- nrow(bmt)
entrytime <- rbinom(nn,1,0.5)*(bmt$time*runif(nn))
bmt$entrytime <- entrytime
times <- seq(5,70,by=1)

### adds weights to uncensored observations
bmtw <- prep.comp.risk(bmt,times=times,time="time",
		       entrytime="entrytime",cause="cause")

#########################################
### nonparametric estimates
#########################################
## {{{ 
### nonparametric estimates, right-censoring only 
out <- comp.risk(Event(time,cause)~+1,data=bmt,
		 cause=1,model="rcif2",
		 times=c(5,30,70),n.sim=0)
out$cum
### same as 
###out <- prodlim(Hist(time,cause)~+1,data=bmt)
###summary(out,cause="1",times=c(5,30,70))

### with truncation 
out <- comp.risk(Event(time,cause)~+1,data=bmtw,cause=1,
  model="rcif2",
  cens.weight=bmtw$cw,weights=bmtw$weights,times=c(5,30,70),
  n.sim=0)
out$cum
### same as
###out <- prodlim(Hist(entry=entrytime,time,cause)~+1,data=bmt)
###summary(out,cause="1",times=c(5,30,70))
## }}} 

#########################################
### Regression 
#########################################
## {{{ 
### with truncation correction
out <- comp.risk(Event(time,cause)~const(tcell)+const(platelet),data=bmtw,
 cause=1,cens.weight=bmtw$cw,
 weights=bmtw$weights,times=times,n.sim=0)
summary(out)

### with only righ-censoring, standard call
outn <- comp.risk(Event(time,cause)~const(tcell)+const(platelet),data=bmt,
	  cause=1,times=times,n.sim=0)
summary(outn)
## }}}

Prints call

Description

Prints call for object. Lists nonparametric and parametric terms of model

Usage

## S3 method for class 'aalen'
print(x, ...)

Arguments

x

an aalen object
...

unused arguments - for S3 compatibility

Author(s)

Thomas Scheike

Identifies the multiplicative terms in Cox-Aalen model and proportional excess risk model

Description

Specifies which of the regressors that belong to the multiplicative part of the Cox-Aalen model

Usage

prop(x)

Arguments

x

variable

Details

λi(t)=Yi(t)(XiT(t)α(t))exp⁡(ZiT(t)β)\lambda_{i}(t) = Y_i(t) ( X_{i}^T(t) \alpha(t) ) \exp(Z_{i}^T(t) \beta )

for this model prop specified the covariates to be included in Zi(t)Z_{i}(t)

Author(s)

Thomas Scheike

Fits Proportional excess hazards model

Description

Fits proportional excess hazards model.

Usage

prop.excess(
  formula = formula(data),
  data = parent.frame(),
  excess = 1,
  tol = 1e-04,
  max.time = NULL,
  n.sim = 1000,
  alpha = 1,
  frac = 1
)

Arguments

formula

a formula object, with the response on the left of a ‘~’ operator, and the terms on the right. The response must be a survival object as returned by the ‘Surv’ function.
data

a data.frame with the variables.
excess

specifies for which of the subjects the excess term is present. Default is that the term is present for all subjects.
tol

tolerance for numerical procedure.
max.time

stopping considered time-period if different from 0. Estimates thus computed from [0,max.time] if max.time>0. Default is max of data.
n.sim

number of simulations in re-sampling.
alpha

tuning paramter in Newton-Raphson procedure. Value smaller than one may give more stable convergence.
frac

number between 0 and 1. Is used in supremum test where observed jump times t1, ..., tk is replaced by t1, ..., tl with l=round(frac*k).

Details

The models are written using the survival modelling given in the survival package.

The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Value

Returns an object of type "prop.excess". With the following arguments:

cum

estimated cumulative regression functions. First column contains the jump times, then follows the estimated components of additive part of model and finally the excess cumulative baseline.

var.cum

robust pointwise variance estimates for estimated cumulatives.

gamma

estimate of parametric components of model.

var.gamma

robust variance estimate for gamma.

pval

p-value of Kolmogorov-Smirnov test (variance weighted) for excess baseline and Aalen terms, H: B(t)=0.

pval.HW

p-value of supremum test (corresponding to Hall-Wellner band) for excess baseline and Aalen terms, H: B(t)=0. Reported in summary.

pval.CM

p-value of Cramer von Mises test for excess baseline and Aalen terms, H: B(t)=0.

quant

95 percent quantile in distribution of resampled Kolmogorov-Smirnov test statistics for excess baseline and Aalen terms. Used to construct 95 percent simulation band.

quant95HW

95 percent quantile in distribution of resampled supremum test statistics corresponding to Hall-Wellner band for excess baseline and Aalen terms. Used to construct 95 percent Hall-Wellner band.

simScoreProp

observed scoreprocess and 50 resampled scoreprocesses (under model). List with 51 elements.

Author(s)

Torben Martinussen

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer Verlag (2006).

Examples

###working on memory leak issue, 3/3-2015
###data(melanoma)
###lt<-log(melanoma$thick)          # log-thickness 
###excess<-(melanoma$thick>=210)    # excess risk for thick tumors
###
#### Fits Proportional Excess hazards model 
###fit<-prop.excess(Surv(days/365,status==1)~sex+ulc+cox(sex)+
###                 cox(ulc)+cox(lt),melanoma,excess=excess,n.sim=100)
###summary(fit)
###par(mfrow=c(2,3))
###plot(fit)

Fit Semiparametric Proportional 0dds Model

Description

Fits a semiparametric proportional odds model:

logit(1−SZ(t))=log(G(t))+βTZlogit(1-S_Z(t)) = log(G(t)) + \beta^T Z

where G(t) is increasing but otherwise unspecified. Model is fitted by maximising the modified partial likelihood. A goodness-of-fit test by considering the score functions is also computed by resampling methods.

Usage

prop.odds(
  formula,
  data = parent.frame(),
  beta = NULL,
  Nit = 20,
  detail = 0,
  start.time = 0,
  max.time = NULL,
  id = NULL,
  n.sim = 500,
  weighted.test = 0,
  profile = 1,
  sym = 0,
  baselinevar = 1,
  clusters = NULL,
  max.clust = 1000,
  weights = NULL
)

Arguments

formula

a formula object, with the response on the left of a '~' operator, and the terms on the right. The response must be a Event object as returned by the ‘Event’ function.
data

a data.frame with the variables.
beta

starting value for relative risk estimates
Nit

number of iterations for Newton-Raphson algorithm.
detail

if 0 no details is printed during iterations, if 1 details are given.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. This is very useful to obtain stable estimates, especially for the baseline. Default is max of data.
id

For timevarying covariates the variable must associate each record with the id of a subject.
n.sim

number of simulations in resampling.
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
profile

if profile is 1 then modified partial likelihood is used, profile=0 fits by simple estimating equation. The modified partial likelihood is recommended.
sym

to use symmetrized second derivative in the case of the estimating equation approach (profile=0). This may improve the numerical performance.
baselinevar

set to 0 to omit calculations of baseline variance.
clusters

to compute cluster based standard errors.
max.clust

number of maximum clusters to be used, to save time in iid decomposition.
weights

weights for score equations.

Details

The modelling formula uses the standard survival modelling given in the survival package.

For large data sets use the divide.conquer.timereg of the mets package to run the model on splits of the data, or the alternative estimator by the cox.aalen function.

The data for a subject is presented as multiple rows or "observations", each of which applies to an interval of observation (start, stop]. The program essentially assumes no ties, and if such are present a little random noise is added to break the ties.

Value

returns an object of type 'cox.aalen'. With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

the martingale based pointwise variance estimates.

robvar.cum

robust pointwise variances estimates.

gamma

estimate of proportional odds parameters of model.
var.gamma

variance for gamma.

robvar.gamma

robust variance for gamma.

residuals

list with residuals. Estimated martingale increments (dM) and corresponding time vector (time).
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.

test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect over time.
loglike

modified partial likelihood, pseudo profile likelihood for regression parameters.
D2linv

inverse of the derivative of the score function.
score

value of score for final estimates.
test.procProp

observed score process for proportional odds regression effects.
pval.Prop

p-value based on resampling.
sim.supProp

re-sampled supremum values.
sim.test.procProp

list of 50 random realizations of test-processes for constant proportional odds under the model based on resampling.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

data(sTRACE)
# Fits Proportional odds model 
out<-prop.odds(Event(time,status==9)~age+diabetes+chf+vf+sex,
sTRACE,max.time=7,n.sim=100)
summary(out)

par(mfrow=c(2,3))
plot(out,sim.ci=2)
plot(out,score=1) 

pout <- predict(out,Z=c(70,0,0,0,0))
plot(pout)

### alternative estimator for large data sets 
form <- Surv(time,status==9)~age+diabetes+chf+vf+sex
pform <- timereg.formula(form)
out2<-cox.aalen(pform,data=sTRACE,max.time=7,
	propodds=1,n.sim=0,robust=0,detail=0,Nit=40)
summary(out2)

Fit Semiparametric Proportional 0dds Model for the competing risks subdistribution

Description

Fits a semiparametric proportional odds model:

logit(F1(t;X,Z))=log(A(t))+βTZlogit(F_1(t;X,Z)) = log( A(t)) + \beta^T Z

where A(t) is increasing but otherwise unspecified. Model is fitted by maximising the modified partial likelihood. A goodness-of-fit test by considering the score functions is also computed by resampling methods.

Usage

prop.odds.subdist(
  formula,
  data = parent.frame(),
  cause = 1,
  beta = NULL,
  Nit = 10,
  detail = 0,
  start.time = 0,
  max.time = NULL,
  id = NULL,
  n.sim = 500,
  weighted.test = 0,
  profile = 1,
  sym = 0,
  cens.model = "KM",
  cens.formula = NULL,
  clusters = NULL,
  max.clust = 1000,
  baselinevar = 1,
  weights = NULL,
  cens.weights = NULL
)

Arguments

formula

a formula object, with the response on the left of a '~' operator, and the terms on the right. The response must be an object as returned by the ‘Event’ function.
data

a data.frame with the variables.
cause

cause indicator for competing risks.
beta

starting value for relative risk estimates
Nit

number of iterations for Newton-Raphson algorithm.
detail

if 0 no details is printed during iterations, if 1 details are given.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. This is very useful to obtain stable estimates, especially for the baseline. Default is max of data.
id

For timevarying covariates the variable must associate each record with the id of a subject.
n.sim

number of simulations in resampling.
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
profile

use profile version of score equations.
sym

to use symmetrized second derivative in the case of the estimating equation approach (profile=0). This may improve the numerical performance.
cens.model

specifies censoring model. So far only Kaplan-Meier "KM".
cens.formula

possible formula for censoring distribution covariates. Default all !
clusters

to compute cluster based standard errors.
max.clust

number of maximum clusters to be used, to save time in iid decomposition.
baselinevar

set to 0 to save time on computations.
weights

additional weights.
cens.weights

specify censoring weights related to the observations.

Details

An alternative way of writing the model :

F1(t;X,Z)=exp⁡(βTZ)(A(t))+exp⁡(βTZ)F_1(t;X,Z) = \frac{ \exp( \beta^T Z )}{ (A(t)) + \exp( \beta^T Z) }

such that β\beta is the log-odds-ratio of cause 1 before time t, and A(t)A(t) is the odds-ratio.

The modelling formula uses the standard survival modelling given in the survival package.

The data for a subject is presented as multiple rows or "observations", each of which applies to an interval of observation (start, stop]. The program essentially assumes no ties, and if such are present a little random noise is added to break the ties.

Value

returns an object of type 'cox.aalen'. With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

the martingale based pointwise variance estimates.

robvar.cum

robust pointwise variances estimates.

gamma

estimate of proportional odds parameters of model.
var.gamma

variance for gamma.

robvar.gamma

robust variance for gamma.

residuals

list with residuals. Estimated martingale increments (dM) and corresponding time vector (time).
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.

test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect over time.
loglike

modified partial likelihood, pseudo profile likelihood for regression parameters.
D2linv

inverse of the derivative of the score function.
score

value of score for final estimates.
test.procProp

observed score process for proportional odds regression effects.
pval.Prop

p-value based on resampling.
sim.supProp

re-sampled supremum values.
sim.test.procProp

list of 50 random realizations of test-processes for constant proportional odds under the model based on resampling.

Author(s)

Thomas Scheike

References

Eriksson, Li, Zhang and Scheike (2014), The proportional odds cumulative incidence model for competing risks, Biometrics, to appear.

Scheike, A flexible semiparametric transformation model for survival data, Lifetime Data Anal. (2007).

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

library(timereg)
data(bmt)
# Fits Proportional odds model 
out <- prop.odds.subdist(Event(time,cause)~platelet+age+tcell,data=bmt,
 cause=1,cens.model="KM",detail=0,n.sim=1000)
summary(out) 
par(mfrow=c(2,3))
plot(out,sim.ci=2); 
plot(out,score=1) 

# simple predict function without confidence calculations 
pout <- predictpropodds(out,X=model.matrix(~platelet+age+tcell,data=bmt)[,-1])
matplot(pout$time,pout$pred,type="l")

# predict function with confidence intervals
pout2 <- predict(out,Z=c(1,0,1))
plot(pout2,col=2)
pout1 <- predictpropodds(out,X=c(1,0,1))
lines(pout1$time,pout1$pred,type="l")

# Fits Proportional odds model with stratified baseline, does not work yet!
###out <- Gprop.odds.subdist(Surv(time,cause==1)~-1+factor(platelet)+
###prop(age)+prop(tcell),data=bmt,cause=bmt$cause,
###cens.code=0,cens.model="KM",causeS=1,detail=0,n.sim=1000)
###summary(out) 
###par(mfrow=c(2,3))
###plot(out,sim.ci=2); 
###plot(out,score=1)

For internal use

Description

for internal use

Author(s)

Thomas Scheike

Cut a variable

Description

Calls the cut function to cut variables on data frame.

Usage

qcut(x, cuts = 4, breaks = NULL, ...)

Arguments

x

variable to cut
cuts

number of groups, 4 gives quartiles
breaks

can also give breaks
...

other argument for cut function of R

Author(s)

Thomas Scheike

Examples

data(sTRACE)
gx <- qcut(sTRACE$age)
table(gx)

Estimates marginal mean of recurrent events based on two cox models

Description

Fitting two Cox models for death and recurent events these are combined to prducte the estimator

∫0tS(u∣x=0)dR(u∣x=0)\int_0^t S(u|x=0) dR(u|x=0)

the mean number of recurrent events, here

S(u∣x=0)S(u|x=0)

is the probability of survival, and

dR(u∣x=0)dR(u|x=0)

is the probability of an event among survivors. For now the estimator is based on the two-baselines so

x=0x=0

, but covariates can be rescaled to look at different x's and extensions possible.

Usage

recurrent.marginal.coxmean(recurrent, death)

Arguments

recurrent

aalen model for recurrent events
death

cox.aalen (cox) model for death events

Details

IID versions along the lines of Ghosh & Lin (2000) variance. See also mets package for quick version of this for large data. IID versions used for Ghosh & Lin (2000) variance. See also mets package for quick version of this for large data mets:::recurrent.marginal, these two version should give the same when there are now ties.

Author(s)

Thomas Scheike

References

Ghosh and Lin (2002) Nonparametric Analysis of Recurrent events and death, Biometrics, 554–562.

Examples

### do not test because iid slow  and uses data from mets
library(mets)
data(base1cumhaz)
data(base4cumhaz)
data(drcumhaz)
dr <- drcumhaz
base1 <- base1cumhaz
base4 <- base4cumhaz
rr <- simRecurrent(100,base1,death.cumhaz=dr)
rr$x <- rnorm(nrow(rr)) 
rr$strata <- floor((rr$id-0.01)/50)
drename(rr) <- start+stop~entry+time

ar <- cox.aalen(Surv(start,stop,status)~+1+prop(x)+cluster(id),data=rr,
                   resample.iid=1,,max.clust=NULL,max.timepoint.sim=NULL)
ad <- cox.aalen(Surv(start,stop,death)~+1+prop(x)+cluster(id),data=rr,
                   resample.iid=1,,max.clust=NULL,max.timepoint.sim=NULL)
mm <- recurrent.marginal.coxmean(ar,ad)
with(mm,plot(times,mu,type="s"))
with(mm,lines(times,mu+1.96*se.mu,type="s",lty=2))
with(mm,lines(times,mu-1.96*se.mu,type="s",lty=2))

Estimates marginal mean of recurrent events

Description

Fitting two aalen models for death and recurent events these are combined to prducte the estimator

∫0tS(u)dR(u)\int_0^t S(u) dR(u)

the mean number of recurrent events, here

S(u)S(u)

is the probability of survival, and

dR(u)dR(u)

is the probability of an event among survivors.

Usage

recurrent.marginal.mean(recurrent, death)

Arguments

recurrent

aalen model for recurrent events
death

aalen model for recurrent events

Details

IID versions used for Ghosh & Lin (2000) variance. See also mets package for quick version of this for large data mets:::recurrent.marginal, these two version should give the same when there are no ties.

Author(s)

Thomas Scheike

References

Ghosh and Lin (2002) Nonparametric Analysis of Recurrent events and death, Biometrics, 554–562.

Examples

### get some data using mets simulaitons, and avoid dependency, see mets
# library(mets)
# data(base1cumhaz)
# data(base4cumhaz)
# data(drcumhaz)
# dr <- drcumhaz
# base1 <- base1cumhaz
# base4 <- base4cumhaz
# rr <- simRecurrent(100,base1,death.cumhaz=dr)
# rr$x <- rnorm(nrow(rr)) 
# rr$strata <- floor((rr$id-0.01)/50)
# drename(rr) <- start+stop~entry+time
# 
# ar <- aalen(Surv(start,stop,status)~+1+cluster(id),data=rr,resample.iid=1
#                                                      ,max.clust=NULL)
# ad <- aalen(Surv(start,stop,death)~+1+cluster(id),data=rr,resample.iid=1,
#                                                      ,max.clust=NULL)
# mm <- recurrent.marginal.mean(ar,ad)
# with(mm,plot(times,mu,type="s"))
# with(mm,lines(times,mu+1.96*se.mu,type="s",lty=2))
# with(mm,lines(times,mu-1.96*se.mu,type="s",lty=2))

Residual mean life (restricted)

Description

Fits a semiparametric model for the residual life (estimator=1):

E(min⁡(Y,τ)−t∣Y>=t)=h1(g(t,x,z))E( \min(Y,\tau) -t | Y>=t) = h_1( g(t,x,z) )

or cause specific years lost of Andersen (2012) (estimator=3)

E(τ−min⁡(Yj,τ)∣Y>=0)=∫0t(1−Fj(s))ds=h2(g(t,x,z))E( \tau- \min(Y_j,\tau) | Y>=0) = \int_0^t (1-F_j(s)) ds = h_2( g(t,x,z) )

where Yj=∑jYI(ϵ=j)+∞∗I(ϵ=0)Y_j = \sum_j Y I(\epsilon=j) + \infty * I(\epsilon=0) or (estimator=2)

E(τ−min⁡(Yj,τ)∣Y<τ,ϵ=j)=h3(g(t,x,z))=h2(g(t,x,z))Fj(τ,x,z)E( \tau- \min(Y_j,\tau) | Y<\tau, \epsilon=j) = h_3( g(t,x,z) ) = h_2(g(t,x,z)) F_j(\tau,x,z)

where Fj(s,x,z)=P(Y<τ,ϵ=j∣x,z)F_j(s,x,z) = P(Y<\tau, \epsilon=j | x,z ) for a known link-function h()h() and known prediction-function g(t,x,z)g(t,x,z)

Usage

res.mean(
  formula,
  data = parent.frame(),
  cause = 1,
  restricted = NULL,
  times = NULL,
  Nit = 50,
  clusters = NULL,
  gamma = 0,
  n.sim = 0,
  weighted = 0,
  model = "additive",
  detail = 0,
  interval = 0.01,
  resample.iid = 1,
  cens.model = "KM",
  cens.formula = NULL,
  time.pow = NULL,
  time.pow.test = NULL,
  silent = 1,
  conv = 1e-06,
  estimator = 1,
  cens.weights = NULL,
  conservative = 1,
  weights = NULL
)

Arguments

formula

a formula object, with the response on the left of a '~' operator, and the terms on the right. The response must be a survival object as returned by the ‘Event’ function. The status indicator is not important here. Time-invariant regressors are specified by the wrapper const(), and cluster variables (for computing robust variances) by the wrapper cluster().
data

a data.frame with the variables.
cause

For competing risk models specificies which cause we consider.
restricted

gives a possible restriction times for means.
times

specifies the times at which the estimator is considered. Defaults to all the times where an event of interest occurs, with the first 10 percent or max 20 jump points removed for numerical stability in simulations.
Nit

number of iterations for Newton-Raphson algorithm.
clusters

specifies cluster structure, for backwards compability.
gamma

starting value for constant effects.
n.sim

number of simulations in resampling.
weighted

Not implemented. To compute a variance weighted version of the test-processes used for testing time-varying effects.
model

"additive", "prop"ortional.
detail

if 0 no details are printed during iterations, if 1 details are given.
interval

specifies that we only consider timepoints where the Kaplan-Meier of the censoring distribution is larger than this value.
resample.iid

to return the iid decomposition, that can be used to construct confidence bands for predictions
cens.model

specified which model to use for the ICPW, KM is Kaplan-Meier alternatively it may be "cox" or "aalen" model for further flexibility.
cens.formula

specifies the regression terms used for the regression model for chosen regression model. When cens.model is specified, the default is to use the same design as specified for the competing risks model. "KM","cox","aalen","weights". "weights" are user specified weights given is cens.weight argument.
time.pow

specifies that the power at which the time-arguments is transformed, for each of the arguments of the const() terms, default is 1 for the additive model and 0 for the proportional model.
time.pow.test

specifies that the power the time-arguments is transformed for each of the arguments of the non-const() terms. This is relevant for testing if a coefficient function is consistent with the specified form A_l(t)=beta_l t^time.pow.test(l). Default is 1 for the additive model and 0 for the proportional model.
silent

if 0 information on convergence problems due to non-invertible derviates of scores are printed.
conv

gives convergence criterie in terms of sum of absolute change of parameters of model
estimator

specifies what that is estimated.
cens.weights

censoring weights for estimating equations.
conservative

for slightly conservative standard errors.
weights

weights for estimating equations.

Details

Uses the IPCW for the score equations based on

w(t)Δ(τ)/P(Δ(τ)=1∣T,ϵ,X,Z)(Y(t)−h1(t,X,Z))w(t) \Delta(\tau)/P(\Delta(\tau)=1| T,\epsilon,X,Z) ( Y(t) - h_1(t,X,Z))

and where Δ(τ)\Delta(\tau) is the at-risk indicator given data and requires a IPCW model.

Since timereg version 1.8.4. the response must be specified with the Event function instead of the Surv function and the arguments.

Value

returns an object of type 'comprisk'. With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

pointwise variances estimates.

gamma

estimate of proportional odds parameters of model.
var.gamma

variance for gamma.

score

sum of absolute value of scores.
gamma2

estimate of constant effects based on the non-parametric estimate. Used for testing of constant effects.
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
conf.band

resampling based constant to construct 95% uniform confidence bands.
B.iid

list of iid decomposition of non-parametric effects.
gamma.iid

matrix of iid decomposition of parametric effects.
test.procBeqC

observed test process for testing of time-varying effects
sim.test.procBeqC

50 resample processes for for testing of time-varying effects
conv

information on convergence for time points used for estimation.

Author(s)

Thomas Scheike

References

Andersen (2013), Decomposition of number of years lost according to causes of death, Statistics in Medicine, 5278-5285.

Scheike, and Cortese (2015), Regression Modelling of Cause Specific Years Lost,

Scheike, Cortese and Holmboe (2015), Regression Modelling of Restricted Residual Mean with Delayed Entry,

Examples

data(bmt); 
tau <- 100 

### residual restricted mean life
out<-res.mean(Event(time,cause>=1)~factor(tcell)+factor(platelet),data=bmt,cause=1,
	      times=0,restricted=tau,n.sim=0,model="additive",estimator=1); 
summary(out)

out<-res.mean(Event(time,cause>=1)~factor(tcell)+factor(platelet),data=bmt,cause=1,
	      times=seq(0,90,5),restricted=tau,n.sim=0,model="additive",estimator=1); 
par(mfrow=c(1,3))
plot(out)

### restricted years lost given death
out21<-res.mean(Event(time,cause)~factor(tcell)+factor(platelet),data=bmt,cause=1,
	      times=0,restricted=tau,n.sim=0,model="additive",estimator=2); 
summary(out21)
out22<-res.mean(Event(time,cause)~factor(tcell)+factor(platelet),data=bmt,cause=2,
	      times=0,restricted=tau,n.sim=0,model="additive",estimator=2); 
summary(out22)


### total restricted years lost 
out31<-res.mean(Event(time,cause)~factor(tcell)+factor(platelet),data=bmt,cause=1,
	      times=0,restricted=tau,n.sim=0,model="additive",estimator=3); 
summary(out31)
out32<-res.mean(Event(time,cause)~factor(tcell)+factor(platelet),data=bmt,cause=2,
	      times=0,restricted=tau,n.sim=0,model="additive",estimator=3); 
summary(out32)


### delayed entry 
nn <- nrow(bmt)
entrytime <- rbinom(nn,1,0.5)*(bmt$time*runif(nn))
bmt$entrytime <- entrytime

bmtw <- prep.comp.risk(bmt,times=tau,time="time",entrytime="entrytime",cause="cause")

out<-res.mean(Event(time,cause>=1)~factor(tcell)+factor(platelet),data=bmtw,cause=1,
	      times=0,restricted=tau,n.sim=0,model="additive",estimator=1,
              cens.model="weights",weights=bmtw$cw,cens.weights=1/bmtw$weights); 
summary(out)

Estimates restricted residual mean for Cox or Aalen model

Description

The restricted means are the

∫0τS(t)dt\int_0^\tau S(t) dt

the standard errors are computed using the i.i.d. decompositions from the cox.aalen (that must be called with the argument "max.timpoint.sim=NULL") or aalen function.

Usage

restricted.residual.mean(out, x = 0, tau = 10, iid = 0)

Arguments

out

an "cox.aalen" with a Cox model or an "aalen" model.
x

matrix with covariates for Cox model or additive hazards model (aalen).
tau

restricted residual mean.
iid

if iid=1 then uses iid decomposition for estimation of standard errors.

Details

must have computed iid decomposition of survival models for standard errors to be computed. Note that competing risks models can be fitted but then the interpretation is not clear.

Value

Returns an object. With the following arguments:

mean

restricted mean for different covariates.
var.mean

variance matrix.
se

standard errors.
S0tau

estimated survival functions on time-range [0,tau].
timetau

vector of time arguments for S0tau.

Author(s)

Thomas Scheike

References

D. M. Zucker, Restricted mean life with covariates: Modification and extension of a useful survival analysis method, J. Amer. Statist. Assoc. vol. 93 pp. 702-709, 1998.

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

### this example runs slowly and is therefore donttest
data(sTRACE)
sTRACE$cage <- scale(sTRACE$age)
# Fits Cox model  and aalen model 
out<-cox.aalen(Surv(time,status>=1)~prop(sex)+prop(diabetes)+prop(chf)+
	       prop(vf),data=sTRACE,max.timepoint.sim=NULL,resample.iid=1)
outa<-aalen(Surv(time,status>=1)~sex+diabetes+chf+vf,
data=sTRACE,resample.iid=1)

coxrm <- restricted.residual.mean(out,tau=7,
   x=rbind(c(0,0,0,0),c(0,0,1,0),c(0,0,1,1),c(0,0,0,1)),iid=1)
plot(coxrm)
summary(coxrm)

### aalen model not optimal here 
aalenrm <- restricted.residual.mean(outa,tau=7,
   x=rbind(c(1,0,0,0,0),c(1,0,0,1,0),c(1,0,0,1,1),c(1,0,0,0,1)),iid=1)
with(aalenrm,matlines(timetau,S0tau,type="s",ylim=c(0,1)))
legend("bottomleft",c("baseline","+chf","+chf+vf","+vf"),col=1:4,lty=1)
summary(aalenrm)

mm <-cbind(coxrm$mean,coxrm$se,aalenrm$mean,aalenrm$se)
colnames(mm)<-c("cox-res-mean","se","aalen-res-mean","se")
rownames(mm)<-c("baseline","+chf","+chf+vf","+vf")
mm

Prints summary statistics

Description

Computes p-values for test of significance for nonparametric terms of model, p-values for test of constant effects based on both supremum and integrated squared difference.

Usage

## S3 method for class 'aalen'
summary(object, digits = 3, ...)

Arguments

object

an aalen object.
digits

number of digits in printouts.
...

unused arguments - for S3 compatibility

Details

Returns parameter estimates and their standard errors.

Author(s)

Thomas Scheike

References

Martinussen and Scheike,

Examples

### see help(aalen)

Prints summary statistics for goodness-of-fit tests based on cumulative residuals

Description

Computes p-values for extreme behaviour relative to the model of various cumulative residual processes.

Usage

## S3 method for class 'cum.residuals'
summary(object, digits = 3, ...)

Arguments

object

output from the cum.residuals() function.
digits

number of digits in printouts.
...

unused arguments - for S3 compatibility

Author(s)

Thomas Scheike

Examples

# see cum.residuals for examples

Fit Cox model with partly timevarying effects.

Description

Fits proportional hazards model with some effects time-varying and some effects constant. Time dependent variables and counting process data (multiple events per subject) are possible.

Usage

timecox(
  formula = formula(data),
  data,
  weights,
  subset,
  na.action,
  start.time = 0,
  max.time = NULL,
  id = NULL,
  clusters = NULL,
  n.sim = 1000,
  residuals = 0,
  robust = 1,
  Nit = 20,
  bandwidth = 0.5,
  method = "basic",
  weighted.test = 0,
  degree = 1,
  covariance = 0
)

Arguments

formula

a formula object with the response on the left of a '~' operator, and the independent terms on the right as regressors. The response must be a survival object as returned by the ‘Surv’ function. Time-invariant regressors are specified by the wrapper const(), and cluster variables (for computing robust variances) by the wrapper cluster().
data

a data.frame with the variables.
weights

for analysis
subset

to subset
na.action

to have na.action
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. Default is max of data.
id

For timevarying covariates the variable must associate each record with the id of a subject.
clusters

cluster variable for computation of robust variances.
n.sim

number of simulations in resampling.
residuals

to returns residuals that can be used for model validation in the function cum.residuals
robust

to compute robust variances and construct processes for resampling. May be set to 0 to save memory.
Nit

number of iterations for score equations.
bandwidth

bandwidth for local iterations. Default is 50 % of the range of the considered observation period.
method

Method for estimation. This refers to different parametrisations of the baseline of the model. Options are "basic" where the baseline is written as λ0(t)=exp⁡(α0(t))\lambda_0(t) = \exp(\alpha_0(t)) or the "breslow" version where the baseline is parametrised as λ0(t)\lambda_0(t).
weighted.test

to compute a variance weighted version of the test-processes used for testing time-varying effects.
degree

gives the degree of the local linear smoothing, that is local smoothing. Possible values are 1 or 2.
covariance

to compute covariance estimates for nonparametric terms rather than just the variances.

Details

Resampling is used for computing p-values for tests of timevarying effects.

The modelling formula uses the standard survival modelling given in the survival package.

The data for a subject is presented as multiple rows or 'observations', each of which applies to an interval of observation (start, stop]. When counting process data with the )start,stop] notation is used, the 'id' variable is needed to identify the records for each subject. The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Value

Returns an object of type "timecox". With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

the martingale based pointwise variance estimates.

robvar.cum

robust pointwise variances estimates.

gamma

estimate of parametric components of model.

var.gamma

variance for gamma.

robvar.gamma

robust variance for gamma.

residuals

list with residuals. Estimated martingale increments (dM) and corresponding time vector (time).
obs.testBeq0

observed absolute value of supremum of cumulative components scaled with the variance.
pval.testBeq0

p-value for covariate effects based on supremum test.
sim.testBeq0

resampled supremum values.
obs.testBeqC

observed absolute value of supremum of difference between observed cumulative process and estimate under null of constant effect.
pval.testBeqC

p-value based on resampling.
sim.testBeqC

resampled supremum values.
obs.testBeqC.is

observed integrated squared differences between observed cumulative and estimate under null of constant effect.
pval.testBeqC.is

p-value based on resampling.
sim.testBeqC.is

resampled supremum values.
conf.band

resampling based constant to construct robust 95% uniform confidence bands.

test.procBeqC

observed test-process of difference between observed cumulative process and estimate under null of constant effect over time.

sim.test.procBeqC

list of 50 random realizations of test-processes under null based on resampling.
schoenfeld.residuals

Schoenfeld residuals are returned for "breslow" parametrisation.

Author(s)

Thomas Scheike

References

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

data(sTRACE)
# Fits time-varying Cox model 
out<-timecox(Surv(time/365,status==9)~age+sex+diabetes+chf+vf,
data=sTRACE,max.time=7,n.sim=100)

summary(out)
par(mfrow=c(2,3))
plot(out)
par(mfrow=c(2,3))
plot(out,score=TRUE)

# Fits semi-parametric time-varying Cox model
out<-timecox(Surv(time/365,status==9)~const(age)+const(sex)+
const(diabetes)+chf+vf,data=sTRACE,max.time=7,n.sim=100)

summary(out)
par(mfrow=c(2,3))
plot(out)

The TRACE study group of myocardial infarction

Description

The TRACE data frame contains 1877 patients and is a subset of a data set consisting of approximately 6000 patients. It contains data relating survival of patients after myocardial infarction to various risk factors.

sTRACE is a subsample consisting of 300 patients.

tTRACE is a subsample consisting of 1000 patients.

Format

This data frame contains the following columns:

id

a numeric vector. Patient code.

status

a numeric vector code. Survival status. 9: dead from myocardial infarction, 0: alive, 7,8: dead from other causes.

time

a numeric vector. Survival time in years.

chf

a numeric vector code. Clinical heart pump failure, 1: present, 0: absent.

diabetes

a numeric vector code. Diabetes, 1: present, 0: absent.

vf

a numeric vector code. Ventricular fibrillation, 1: present, 0: absent.

wmi

a numeric vector. Measure of heart pumping effect based on ultrasound measurements where 2 is normal and 0 is worst.

sex

a numeric vector code. 1: female, 0: male.

age

a numeric vector code. Age of patient.

Source

The TRACE study group.

Jensen, G.V., Torp-Pedersen, C., Hildebrandt, P., Kober, L., F. E. Nielsen, Melchior, T., Joen, T. and P. K. Andersen (1997), Does in-hospital ventricular fibrillation affect prognosis after myocardial infarction?, European Heart Journal 18, 919–924.

Examples

data(TRACE)
names(TRACE)

Fit Clayton-Oakes-Glidden Two-Stage model

Description

Fit Clayton-Oakes-Glidden Two-Stage model with Cox-Aalen marginals and regression on the variance parameters.

Usage

two.stage(
  margsurv,
  data = parent.frame(),
  Nit = 60,
  detail = 0,
  start.time = 0,
  max.time = NULL,
  id = NULL,
  clusters = NULL,
  robust = 1,
  theta = NULL,
  theta.des = NULL,
  var.link = 0,
  step = 0.5,
  notaylor = 0,
  se.clusters = NULL
)

Arguments

margsurv

fit of marginal survival cox.aalen model with residuals=2, and resample.iid=1 to get fully correct standard errors. See notaylor below.
data

a data.frame with the variables.
Nit

number of iterations for Newton-Raphson algorithm.
detail

if 0 no details is printed during iterations, if 1 details are given.
start.time

start of observation period where estimates are computed.
max.time

end of observation period where estimates are computed. Estimates thus computed from [start.time, max.time]. Default is max of data.
id

For timevarying covariates the variable must associate each record with the id of a subject.
clusters

cluster variable for computation of robust variances.
robust

if 0 then totally omits computation of standard errors.
theta

starting values for the frailty variance (default=0.1).
theta.des

design for regression for variances. The defauls is NULL that is equivalent to just one theta and the design with only a baseline.
var.link

default "0" is that the regression design on the variances is without a link, and "1" uses the link function exp.
step

step size for Newton-Raphson.
notaylor

if 1 then ignores variation due to survival model, this is quicker and then resample.iid=0 and residuals=0 is ok for marginal survival model that then is much quicker.
se.clusters

cluster variable for sandwich estimator of variance.

Details

The model specifikatin allows a regression structure on the variance of the random effects, such it is allowed to depend on covariates fixed within clusters

θk=QkTν\theta_{k} = Q_{k}^T \nu

. This is particularly useful to model jointly different groups and to compare their variances.

Fits an Cox-Aalen survival model. Time dependent variables and counting process data (multiple events per subject) are not possible !

The marginal baselines are on the Cox-Aalen form

λki(t)=Yki(t)(XkiT(t)α(t))exp⁡(ZkiTβ)\lambda_{ki}(t) = Y_{ki}(t) ( X_{ki}^T(t) \alpha(t) ) \exp(Z_{ki}^T \beta )

The model thus contains the Cox's regression model and the additive hazards model as special cases. (see cox.aalen function for more on this).

The modelling formula uses the standard survival modelling given in the survival package. Only for right censored survival data.

The data for a subject is presented as multiple rows or 'observations', each of which applies to an interval of observation (start, stop]. For counting process data with the )start,stop] notation is used the 'id' variable is needed to identify the records for each subject. Only one record per subject is allowed in the current implementation for the estimation of theta. The program assumes that there are no ties, and if such are present random noise is added to break the ties.

Left truncation is dealt with. Here the key assumption is that the maginals are correctly estimated and that we have a common truncation time within each cluster.

Value

returns an object of type "two.stage". With the following arguments:

cum

cumulative timevarying regression coefficient estimates are computed within the estimation interval.
var.cum

the martingale based pointwise variance estimates.
robvar.cum

robust pointwise variances estimates.
gamma

estimate of parametric components of model.
var.gamma

variance for gamma.
robvar.gamma

robust variance for gamma.
D2linv

inverse of the derivative of the score function from marginal model.
score

value of score for final estimates.
theta

estimate of Gamma variance for frailty.
var.theta

estimate of variance of theta.
SthetaInv

inverse of derivative of score of theta.
theta.score

score for theta parameters.

Author(s)

Thomas Scheike

References

Glidden (2000), A Two-Stage estimator of the dependence parameter for the Clayton Oakes model.

Martinussen and Scheike, Dynamic Regression Models for Survival Data, Springer (2006).

Examples

library(timereg)
data(diabetes)
# Marginal Cox model  with treat as covariate
marg <- cox.aalen(Surv(time,status)~prop(treat)+prop(adult)+
	  cluster(id),data=diabetes,resample.iid=1)
fit<-two.stage(marg,data=diabetes,theta=1.0,Nit=40)
summary(fit)

# using coxph and giving clusters, but SE wittout cox uncetainty
margph <- coxph(Surv(time,status)~treat,data=diabetes)
fit<-two.stage(margph,data=diabetes,theta=1.0,Nit=40,clusters=diabetes$id)


# Stratification after adult 
theta.des<-model.matrix(~-1+factor(adult),diabetes);
des.t<-model.matrix(~-1+factor(treat),diabetes);
design.treat<-cbind(des.t[,-1]*(diabetes$adult==1),
                    des.t[,-1]*(diabetes$adult==2))

# test for common baselines included here 
marg1<-cox.aalen(Surv(time,status)~-1+factor(adult)+prop(design.treat)+cluster(id),
 data=diabetes,resample.iid=1,Nit=50)

fit.s<-two.stage(marg1,data=diabetes,Nit=40,theta=1,theta.des=theta.des)
summary(fit.s)

# with common baselines  and common treatment effect (although test reject this)
fit.s2<-two.stage(marg,data=diabetes,Nit=40,theta=1,theta.des=theta.des)
summary(fit.s2)

# test for same variance among the two strata
theta.des<-model.matrix(~factor(adult),diabetes);
fit.s3<-two.stage(marg,data=diabetes,Nit=40,theta=1,theta.des=theta.des)
summary(fit.s3)

# to fit model without covariates, use beta.fixed=1 and prop or aalen function
marg <- aalen(Surv(time,status)~+1+cluster(id),
	 data=diabetes,resample.iid=1,n.sim=0)
fita<-two.stage(marg,data=diabetes,theta=0.95,detail=0)
summary(fita)

# same model but se's without variation from marginal model to speed up computations
marg <- aalen(Surv(time,status) ~+1+cluster(id),data=diabetes,
	      resample.iid=0,n.sim=0)
fit<-two.stage(marg,data=diabetes,theta=0.95,detail=0)
summary(fit)

# same model but se's now with fewer time-points for approx of iid decomp of marginal 
# model to speed up computations
marg <- cox.aalen(Surv(time,status) ~+prop(treat)+cluster(id),data=diabetes,
	      resample.iid=1,n.sim=0,max.timepoint.sim=5,beta.fixed=1,beta=0)
fit<-two.stage(marg,data=diabetes,theta=0.95,detail=0)
summary(fit)

Makes wald test

Description

Makes wald test, either by contrast matrix or testing components to 0. Can also specify the regression coefficients and the variance matrix. Also makes confidence intervals of the defined contrasts. Reads coefficientes and variances from timereg and coxph objects.

Usage

wald.test(
  object = NULL,
  coef = NULL,
  Sigma = NULL,
  vcov = NULL,
  contrast,
  coef.null = NULL,
  null = NULL,
  print.coef = TRUE,
  alpha = 0.05
)

Arguments

object

timereg object
coef

estimates from some model
Sigma

variance of estimates
vcov

same as Sigma but more standard in other functions
contrast

contrast matrix for testing
coef.null

which indeces to test to 0
null

mean of null, 0 by default
print.coef

print the coefficients of the linear combinations.
alpha

significance level for CI for linear combinations of coefficients.

Author(s)

Thomas Scheike

Examples

data(sTRACE)
# Fits Cox model 
out<-cox.aalen(Surv(time,status==9)~prop(age)+prop(sex)+
prop(vf)+prop(chf)+prop(diabetes),data=sTRACE,n.sim=0)

wald.test(out,coef.null=c(1,2,3))
### test age=sex   vf=chf
wald.test(out,contrast=rbind(c(1,-1,0,0,0),c(0,0,1,-1,0)))

### now same with direct specifation of estimates and variance
wald.test(coef=out$gamma,Sigma=out$var.gamma,coef.null=c(1,2,3))
wald.test(coef=out$gamma,Sigma=out$robvar.gamma,coef.null=c(1,2,3))
### test age=sex   vf=chf
wald.test(coef=out$gamma,Sigma=out$var.gamma,
	  contrast=rbind(c(1,-1,0,0,0),c(0,0,1,-1,0)))