Testing log-linearity


New Member
Hi all,

I want to test the log-linearity assumption for a cox model.
How to interpret the plot of martingale residuals? you need to have a horizontal plot, or a linear line, or ...
In which case should the hypothesis be validated and in which case should it not be validated?


New Member
Yeah that sounds right, horizontal projection of residuals. Post your figure.
Ok thanks if it's horizontal it's validated.
For exemple, if is it a curve that goes up and down or the opposite?
How to interpret it? What to do? Should we reject the hypothesis?


Less is more. Stay pure. Stay poor.
Well nothing is perfect, think about a Q-Q plot. So given it isn't out of control it is likely assumed horizontal.