`sim.fmsm` gives different number of transitions when simulating same sample size

[mafed]

Simulating the same number of subjects from the bosms3 dataset doesn't give the same number of transitions.
I wasn't expecting exactly the same numbers but these differences seem weird.
Here below you have an MRE: I used both tidy_sim.fmsm() (defined in #163) and just looked at number of first transitions.
Might this be related to #153?

### MRE sim.fmsm 
library(tidyverse)
library(flexsurv)

blnorm <- flexsurvreg(Surv(years, status) ~ trans, data = bosms3, 
                      dist = "lnorm")
tmat <- rbind(c(NA,1,2),c(NA,NA,3),c(NA,NA,NA))
set.seed(777)
invisible(rnorm(1e3))
sim_blnorm <- sim.fmsm(bexp, M = nrow(bosms3), t = max(bosms3$years), 
                       trans = tmat) 
sim_blnorm_tidy <- 
  sim_blnorm %>% 
  tidy_sim.fmsm()

bosms3 %>% count(status, trans)
# An object of class 'msdata'
# 
# Data:
#   status trans   n
# 1      0     1 *101*
# 2      0     2 175
# 3      0     3  35
# 4      1     1 *103*
# 5      1     2  29
# 6      1     3  68

sim_blnorm_tidy %>% count(status, trans_num)
# # A tibble: 6 × 3
#   status trans_num     n
#    <int>     <dbl> <int>
# 1      0         1    *56*
# 2      0         2    56
# 3      0         3    81
# 4      1         1   *345*
# 5      1         2   110
# 6      1         3   264

sim_blnorm$st %>% as_tibble() %>% count(V1 == 1, V2 == 2)
# # A tibble: 2 × 3
#   `V1 == 1` `V2 == 2`     n
#   <lgl>     <lgl>     <int>
# 1 TRUE      FALSE       166
# 2 TRUE      TRUE        *345*

[chjackson]

I'm having trouble understanding this report, I'm afraid!

Firstly M isn't the number of subjects in bosms3, it's the number of potential transitions, including competing risks that weren't observed.

Then it is unclear to me what you expect the results should be, because these depend on the function tidy_sim.fmsm that you have written, that's not in the package, and I'm not familiar with what this function is supposed to do.

If you suspect flexsurv is doing something different from what it is documented to do, please could you try to identify the problem more precisely?