martingaleResid {discSurv}R Documentation

Martingale Residuals

Description

Estimates the martingale residuals of discrete survival model.

Usage

martingaleResid(hazards, dataSetLong)

## S3 method for class 'discSurvMartingaleResid'
plot(x, covariates, dataSetLong, ...)

Arguments

hazards

Predicted hazards from a discrete survival model ("numeric vector").

dataSetLong

Data in long format ("class data.frame").

x

Object of class "discSurvMartingaleResid"("class discSurvMartingaleResid")

covariates

Names of covariates to plot ("character vector").

...

Additional arguments to the plot function

Details

Gives a different plot of each marginal covariate against the martingale residuals. Additionally a nonparametric loess estimation is done.

Value

Martingale residuals for each observation in long format ("numeric vector").

Author(s)

Thomas Welchowski welchow@imbie.meb.uni-bonn.de

References

Tutz G, Schmid M (2016). Modeling discrete time-to-event data. Springer Series in Statistics.

Therneau TM, Grambsch PM, Fleming TR (1990). “Martingale-Based Residuals for Survival Models.” Biometrika, 70, 147-160.

See Also

glm

Examples


# Example with cross validation and unemployment data 
library(Ecdat)
data(UnempDur)
summary(UnempDur$spell)

# Extract subset of data
set.seed(635)
IDsample <- sample(1:dim(UnempDur)[1], 100)
UnempDurSubset <- UnempDur [IDsample, ]

# Conversion to long format
UnempDurSubsetLong <- dataLong(dataShort = UnempDurSubset,
timeColumn = "spell", eventColumn = "censor1")

# Estimate discrete survival continuation ratio model
contModel <- glm(y ~ timeInt + age + logwage, data = UnempDurSubsetLong,
family = binomial(link = "logit"))

# Fit hazards to the data set in long format
hazPreds <- predict(contModel, type = "response")

# Calculate martingale residuals for the unemployment data subset
MartResid <- martingaleResid (hazards = hazPreds, dataSetLong = UnempDurSubsetLong)
MartResid
sum(MartResid)

# Plot martingale residuals vs each covariate in the event interval
# Dotted line represents the loess estimate
plot(MartResid, covariates = c("age", "logwage"), dataSetLong = UnempDurSubsetLong)


[Package discSurv version 2.0.0 Index]