topic Re: Friedman test and Survival Curves in Statistical Procedures

Friedman test and Survival Curves

Daria — Tue, 04 Jun 2013 13:29:59 GMT

Hallo colleagues!

I have some questions about SAS/STAT Procedures:
1) I can’t find in SAS/STAT Multifactor Friedman test ANOVA. I see one-way ANOVA test, is there multifactor(and how can I get it)?
2) I have more than 4 survival curves(hypothesis that there is no differences). I can reject or support hypothesis? But what method do I need to perform what curves have differences?

Many Thanks,

-------------------------------------------------------------------------------------

Balashova

Re: Friedman test and Survival Curves

Reeza — Tue, 04 Jun 2013 15:15:17 GMT

For 2) You can do a pairwise test to see which are truely different. Or you can switch to a parametric method, such as PROC PHREG (Cox regression model) that will provide more details on which levels are different.

Re: Friedman test and Survival Curves

SteveDenham — Tue, 04 Jun 2013 17:58:33 GMT

For 1, I'd like more information, but as a start, you could transform the data to ranks and use PROC GLM. Note that the data must be balanced for this approach to work. In addition, there cannot be any interactions between main effects/factors. With more information, it may be possible to suggest an analysis that is not so restrictive, especially if you are doing Friedman's because of "lack of normality".

Steve Denham