Closed kkmann closed 2 years ago
commenting out
resolves the problem but some form of protection against infinite recursion is probably required
Thanks for the report. I've reworked this pathway detection function now, and it should be working in https://github.com/chjackson/flexsurv-dev/commit/39465114cbf1b9f68b9548e3c10bf4559250462a
Hi Chris,
thanks for the quick reply - I checked it and the pathway detection works for me too now. I then ran
> ppath_fmixmsm(fit, B = 20)
final pathway val lower upper
1 death stable-response-progression-death 0.1666667 0.02346216 0.4203543
2 death stable-response-death 0.1666667 0.02854172 0.4040497
3 death stable-death 0.3333333 0.04255448 0.719964
should the path probabilities not add to one? Maybe my data example is too simple? I thought in the mixture model, artificial censoring events are not required for non-observed transitions, right?
With your example above, I get another row with stable-progression-death
final pathway val lower upper
1 death stable-response-progression-death 0.1666667 0.009222188 0.5191553
2 death stable-response-death 0.1666667 0.002622412 0.4903672
3 death stable-progression-death 0.3333333 0.092747042 0.7017576
4 death stable-death 0.3333333 0.067210043 0.8036272
Sorry, only now got back to checking. I can confirm it's working. Thanks again for the quick fix. Really impressive to see what flexsurv can do now!
Hi,
I get the follwoing unexpected behaviour with
fmixmxm()
.Here is my code
Here is the output from inspecting
fit
:I am confused that the pathways attribute is empty since there are many clear pathways here (s->d; s->r->p->d; s->p->d). If I look at the reduced model with the first two components only, all seems to be fine:
Is this expected behaviour (if so why?)?
Thanks,
Kevin