Programming the statistical procedures from SAS

investigation for significant covariate

Reply
Contributor
Posts: 44

investigation for significant covariate

Hi all,

i have one issue right now, please help me to move forward.

i used logistic model with forward selection method to find out significat covariates. in this case i got weight a significant variable from logistic procedure. then i got a comment that "why weight significant variable. please cross check and let me know the reason"? so in this case how can i cross check why this variable (Weight) significant?

please suggest me how to resolve this problem?

Many Thanks,

Hima

Respected Advisor
Posts: 2,655

Re: investigation for significant covariate

Have you looked at data plots to see how the response variable trends with respect to weight?  Also, since you used an automated selection method, there is a very strong possibility that it got included because there are moderating or mediating variables previously included.  Search this site for links to Flom and Cassell's SUGI article pointing out problems with stepwise variable selection methods.

In the end, "why" questions come down to your subject specific knowledge about the field and your data.  Statistical analysis can help illuminate this, but it cannot answer it.

Steve Denham

Ask a Question
Discussion stats
  • 1 reply
  • 268 views
  • 2 likes
  • 2 in conversation