Skip to content

Problem with mssample in a clock reset approach with history and output "data" #8

Description

@MichaelStefan

Dear Dr. Putter,

I have been using the mstate package for years and am very glad for this great package. However, recently I had a problem when generating the output „data“ with the mssample function in a setting with a „clock reset“ approach and using history. To illustrate my problem, here is some code using a simple illness death model:

Haz <- data.frame(time = rep(seq(0, 9.9, 0.1), 3), Haz = rep(seq(0, 0.99, 0.01), 3), trans = rep(c(1,2,3), each = 100))
trans <- matrix(c(NA, 1, 2, NA, NA, 3, NA, NA, NA), ncol = 3, byrow = T)
alt <- mstate::mssample(Haz = Haz,
trans = trans,
history = list(time = 2, state = 1),
clock = "reset",
M = 100,
output = "data",
do.trace = 100)
table(alt$Tstop[alt$trans==1])
table(alt$duration[alt$trans==1])

It seems that the minimum simulated duration in the first state is the history time. As far as I understand the history setting, a history state 1 and time 2 means that the patient is still in state 1 at 2 time units after entering this state. Thus, it seems to me that there is no reason, why a patient should not enter another state earlier than time 4.

To tackle this problem, I had a look into the code and think I located the problem in the function mssample1. I replaced the line
tcond <- t0 <- Tstart <- history$time
by the lines
tcond <- 0
t0 <- Tstart <- history$time
and think this solved my problem. However, I am still not sure, if I understood anything completely wrong and have of course not tested, how this adjustment will influence other settings.

Thanks in advance and best regards,
Michael

Activity

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

bugSomething isn't working

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions