mets (version 1.3.11)

test_marginalMean: Pepe-Mori Test for Marginal Mean Comparison

Description

Performs score-test type tests for proportionality of marginal means in competing risks data and recurrent events, as presented in Ghosh and Lin (2000). The test is based on an IPCW (Inverse Probability of Censoring Weighting) formulation.

Usage

test_marginalMean(
  formula,
  data,
  cause = 1,
  cens.code = 0,
  ...,
  death.code = 2,
  death.code.prop = NULL,
  time = NULL,
  beta = NULL
)

Value

An object of class "marginalTest" containing:

pepe.mori

Pepe-Mori test results with compare p-value.

RatioAUC

Ratio of AUC test results with compare p-value.

difAUC

Difference of AUC test results with compare p-value.

prop.test

Proportionality test results.

score.test

Score test results (equivalent to Gray's test).

score.iid

Influence function for the score test.

time

Upper time limit used.

RAUCl, RAUCe

Raw and transformed AUC estimates.

Arguments

formula

Formula with an Event object on the left-hand side and covariates (typically with strata() for group comparison) on the right. Can include cluster(id) for correlated data.

data

Data frame containing all variables referenced in the formula.

cause

Cause of interest (default 1).

cens.code

Censoring code (default 0).

...

Additional arguments passed to lower-level functions.

death.code

Code for death (terminating event, default 2).

death.code.prop

Code for other causes of death for Fine-Gray regression model.

time

Upper limit for Pepe-Mori and AUC integrals. If NULL, defaults to the maximum event time for the cause of interest.

beta

Starting values for the score test (default NULL, uses zeros).

Author

Thomas Scheike

Details

The function computes several tests:

1. Pepe-Mori Test: Tests for equality of marginal mean functions between groups.

2. Ratio of AUC: Compares the area under the curve of marginal means.

3. Difference of AUC: Tests for difference in areas under the curve.

4. Score Test: Tests for proportionality (equivalent to Gray's test for CIF).

5. Proportionality Test: Tests the proportional hazards assumption.

The Pepe-Mori test uses weights based on the number at risk in each group to construct a weighted integral of the difference in marginal means.

References

Ghosh, D. and Lin, D. Y. (2000). Nonparametric Analysis of Recurrent Events and Death. Biometrics, 56, 554--562.

See Also

test_logrankRecurrent, test_conc

Examples

data(bmt,package="mets")
bmt$time <- bmt$time+runif(nrow(bmt))*0.01
bmt$id <- 1:nrow(bmt)
dcut(bmt) <- age.f~age
     
fg=cifregFG(Event(time,cause)~tcell,data=bmt,cause=1)
 
## computing tests for difference  for CIF
pmt <- test_marginalMean(Event(time,cause)~strata(tcell)+cluster(id),data=bmt,cause=1,
			 death.code=1:2,death.code.prop=2,cens.code=0,time=40)
summary(pmt) 
 
pmt$pepe.mori
pmt$RatioAUC
pmt$prop.test
## score test equialent to Gray's test but variance estimated differently 
pmt$score.test
 
### age-groups  
pmt <- test_marginalMean(Event(time,cause)~strata(age.f)+cluster(id),data=bmt,cause=1,
			 death.code=1:2,death.code.prop=2,cens.code=0)
summary(pmt) 
 
## having a look at the cumulative incidences 
cifs <- cif(Event(time,cause)~strata(age.f)+cluster(id),data=bmt,cause=1)
plot(cifs) 
 
## recurrent events   
data(hfactioncpx12)
hf <- hfactioncpx12
pmt <- test_marginalMean(Event(entry,time,status)~strata(treatment)+cluster(id),data=hf,
			 cause=1,death.code=2,cens.code=0)
summary(pmt)