Thank you very much Terry. I'm still puzzled at why this worked a year
ago. What changed? I'd very much like to reverse the change by setting
an argument somewhere or manipulating the terms object.
I echo your sentiments about the general approach.
Frank
On 06/15/2015 09:05 AM, Therneau, T
Frank,
I don't think there is any way to "fix" your problem except the way that I
did it.
library(survival)
tdata <- data.frame(y=c(1,3,3,5, 5,7, 7,9, 9,13),
x1=factor(letters[c(1,1,1,1,1,2,2,2,2,2)]),
x2= c(1,2,1,2,1,2,1,2,1,2))
fit1 <- lm( y ~ x1 * st
Terry - your example didn't demonstrate the problem because the variable
that interacted with strata (zed) was not a factor variable.
But I had stated the problem incorrectly. It's not that there are too
many strata terms; there are too many non-strata terms when the variable
interacting with
Frank,
I'm not sure what is going on. The following test function works for me in both 3.1.1
and 3.2, i.e, the second model matrix has fewer columns. As I indicated to you earlier,
the coxph code removes the strata() columns after creating X because I found it easier to
correctly create the
4 matches
Mail list logo