There are a few packages that pcox has as dependencies, for which we're only using 1 or 2 functions.
pryr: the modify_call() function
refund: I think there are only a couple accessory functions that we use? Things like safeDeparse().
coxme: I don't think we are using this at all at the moment. The thought was that eventually we could use coxme() as the "fitter" instead of coxph() - it is designed for ML estimation, instead of the grid search that we are using. But last I checked, coxme() requires a "variance matrix", not a penalty matrix, which is a problem for us if the penalty matrix is not invertible.
It would be nice to eliminate these dependencies. coxme we could just remove for now, but for the others, I'm not sure what the etiquette is. Could I just copy the code for the 1 or 2 small functions that we need, and put them in pcox?
There are a few packages that
pcox
has as dependencies, for which we're only using 1 or 2 functions.pryr
: themodify_call()
functionrefund
: I think there are only a couple accessory functions that we use? Things likesafeDeparse()
.coxme
: I don't think we are using this at all at the moment. The thought was that eventually we could usecoxme()
as the "fitter" instead ofcoxph()
- it is designed for ML estimation, instead of the grid search that we are using. But last I checked,coxme()
requires a "variance matrix", not a penalty matrix, which is a problem for us if the penalty matrix is not invertible.It would be nice to eliminate these dependencies.
coxme
we could just remove for now, but for the others, I'm not sure what the etiquette is. Could I just copy the code for the 1 or 2 small functions that we need, and put them inpcox
?