0

so I'm not great with stats but I'm a little confused about this;

  1. I've performed a multivariate cox regression analysis using various categorical variables regarding overall survival. My variable of interest (I'll call it A') was significant in univariate but no longer significant in multivariate analysis. Whereas other factors I wasn't focused on were significant in both analyses, therefore an independent prognostic marker (I'll call one of these factors B'). So my first question is why might a variable lose significane from univariate to multivariate analysis? I understand it is a viable result, but I don't know how to explain it? Would A' be significantly associated with OS and not independently prognostic, therefore the association with OS is possibly mediated through one of the factors that were independently prognostic (e.g B')? Or is there an alternative way to look at it?

Any advice would be greatly appreciated!

rob_stevens95
  • 1
  • 1
  • Your various questions in (1) have answers already here on CV, so please search our site. Good keywords include "multiple regression," "significan*", "not" (to detect questions about significance and non-significance), "mediation," "control" (as in "control for"). The meaning of (2) is unclear, because you already describe such a test. – whuber Sep 16 '21 at 15:11

0 Answers0