What p value to use for purposeful selection of variables for multivariate

#1
To do purposeful selection of variables for multivariate, can I use a p value of 0.15 cut off to select those significant variables from univariate
Though experts say univariate initially is not necessary, bulk of papers, theses do this first and then select those with p <0.05.

Can I do purposeful selection at p 0.15

Kindly help
 

hlsmith

Omega Contributor
#2
Seems fine with me, but as you mentioned not necessary. It is mostly about using your background knowledge of the subject to select applicable candidate variables without blurring inferential analyses with exploratory!