flexsurv-dev

flexsurv-dev copied to clipboard

`sim.fmsm()` different results with `tidy=TRUE` vs `tidy=FALSE`

2

The MRE below shows that only uncensored paths are returned when `tidy=TRUE`. Is this on purpose @chjackson? I couldn't find reference to this behaviour in the help page. I think...

mafed

New S3 methods for fmsm objects

2

Hi @chjackson, Would you be open to a pull request for a new `predict.fmsm` method for the `predict` S3 generic? I would generally match the argument signature to the `predict.flexsurvreg`...

mattwarkentin

Expected number of visits from a `fmsm()` model

1

Hi @chjackson, Is it possible to estimate the expected number of visits in each state from a `flexsurv::fmsm()` model? Similar to the `envisits.msm()` function from the `{msm}` package. I guess...

mattwarkentin

Sampling survival probabilities

6

Would be useful to have a way to generate a set of survival probabilities based on drawing a new set of parameters from `mvnorm`. This can effectively be done right...

jrdnmdhl

allow conditional sampling in 'sim.fmsm()' ?

4

Hi, this is related to #150 - I am trying to sample forward from the predictive distribution of an individual given current state and sojourn time in that state for...

kkmann

More methods needed for residuals.flexsurvreg

8

The current implementation only supports difference between observed survival and predicted mean, which ignores censoring, so is of limited use. The `survival` package seems to do something more considered for...

chjackson

Using `flexsurv` to fit a piecewise exponential model

4

Could anyone provide some hint if I can fit a piecewise exponential model using `flexsurv`? I feel I can do it using `flexsurvspline`, but not sure how to define the...

elong0527

Numerical underflow in pgengamma

1

`pgengamma` underflows for extreme values of Q: ``` > pgengamma(c(56.58, 56.59), mu=0.7, sigma=exp(-1.5707), Q=-45.9621) [1] 0.2968058 1.0000000 ``` This behaviour is inherited from `pgamma`: ``` > pgamma(c(4e-324, 4e-325), 0.0004, 1)...

chjackson

Flexible parametric competing risks models

1

Hi @chjackson, Do you have any thoughts about whether this package would ever add support for being able to use flexible parametric models to directly model the cause-specific cumulative incidences?...

mattwarkentin

flexsurvreg handles uncentered and/or high-variance predictors particularly badly

2

I have what I'm sure is a very common use case: a predictor that's a timestamp converted to numeric, in seconds. This creates very large values, which often have very...

dwalsh-css