Closed jgellar closed 8 years ago
Also, it would be nice if we could replace the Call:
line at the top with the pcox
call, but I'm not too picky on this point.
So I killed two birds with one stone, and wrote a short print.pcox
method that replaces object$call
with object$pcox$call
and then calls print.coxph
. So now the na.action
function is not an issue. I'd still like your input on issue number 2 above though (what to include in the table for these terms).
Here is the current output when a
pcox
models is printed, which goes throughprint.coxph
:There are two things that aren't ideal:
na.action
function (na.omit_pcox
) gets printed out. Any ideas how to get rid of it? Why does this not happen when one of the standardna.action
functions (e.g.,na.omit
) is used? One thought was because I havena.omit_pcox
as internal, but I don't think that would be the only cause of this - and when I exported the function nothing printed differently.coxph.penalty
terms: you can replace those 9 rows with basically whatever you want. This is done by specifying aprintfun
function as an attribute of thecoxph.penalty
term. For example, in hispspline
function, this is the output:So he replaces his 10 spline coefficients with 2 lines, which he calls
linear
andnonlinear
.linear
I believe would be the coefficient for a simple linear term for age, and non-linear has something to do with the deviation for this. He adds a test in for theChisq
,DF
, andp
columns for each row - though I don't really believe this test at all. I'm guessing it's similar to Wood's test insummary.gam
.Anyways, we can replace our 9 rows with anything, with the caveat that it has to fall into these 6 columns. Blanks are allowed. Our options are:
3 would be ideal if we had a good idea on how to do that. You got anything? We don't have any tests developed yet - I have wanted to think about this at some point down the line.