Stepwise Variable Selection Procedure for survreg
This stepwise variable selection procedure can be applied to obtain the best candidates for a survreg
fit.
srstepwise(x, times, delta, sle = 0.15, sls = 0.15, dist='lognormal')
x
: Matrix of variables to consider.times
: The time to an event, if any.delta
: The event indicator: 1 for event, 0 for no event.sle
: The chosen significance level for entering.sls
: The chosen significance level for staying.dist
: The distribution to be used by survreg
.Returns a list of indices of variables which have entered and stayed.
Unfortunately, no stepwise procedure exists for survreg
models. Therefore, we provide this brute force method.
names. <- names(lung)[-(2:3)] status1 <- ifelse(lung$status==2,1,0) X <- as.matrix(lung)[ , names.] vars=srstepwise(X, lung$time, status1) print(names.[vars])
lung
Useful links