Modifications in Cox model visualization

[DanChaltiel]

Hi Alboukadel,

First, thank you for this great package, your code is incredible!

Here is a little pull request on 2 functions.

ggcoxfunctional.R

On ggcoxfunctional.R, I only made a minor modification which correct an error I get a lot.

Indeed, if a model includes categorical variables, the function will throw :

Error in xy.coords(x, y, setLab = FALSE) : 'x' and 'y' lengths differ

Since every variable is plotted against the null model, I thought it makes sence to just exclude categorical variables.

ggcoxzph.R

On ggcoxzph.R, modifications are quite substantial and maybe subject to discussion.

Your function plots an estimation of beta as a function of time. In order to validate the proportionnal hazard assumption, I find very valuable to plot the beta estimated by the model on top of your graph.

In order to access this coefficient, I need the fit object to be of class coxph so I added support to this class. Then I just had to add a geom_hline call plus some arguments and documentation.

I didn't see any testthat in your package to comply with, but I think I tested these functions pretty thoroughly.

Hope it can help :-)
Dan

[DanChaltiel]

Sorry, I'm not familliar with travis-ci, and its documentation is not really helping.

How can I know where the error is?