Dear List I have few queries in survival analysis. 1. Is it possible to get a significance in Univariate Cox regression for a variable which was not significant in chi square with recurrence as dependent variable? (Chi-square p value 0.808 and the crude RR is 0.030 with good 95%CI) 2. How do we do the model building for multivariate? I took the variables whose p value was less than 0.25 ( not to miss any variable due to confounding) and did enter them together and removed or accepted depending on the improvement in model fit. 3. If a variable Log minus log curve is not parallel, should that be dropped? Thanks Jayabaskar . . ================================================================= Instructions for joining and leaving this list, remarks about the problem of INAPPROPRIATE MESSAGES, and archives are available at: . http://jse.stat.ncsu.edu/ . =================================================================
