Programming the statistical procedures from SAS

Stepwise regression with year effects

Reply
New Contributor
Posts: 2

Stepwise regression with year effects

Hello,

I have a repeated measures ecological data set for which I need to examine individual species-habitat responses based on site type.

There are three site groupings (i.e., recent, old and control) and each site group has five sample sites (i.e., site 1-5, 6-10, and 11-15).

The data were collected at the same sites over two years (i.e.,2012 and 2013), twice per year (i.e., spring and summer).

In total I have four species that I need to analyze for the spring and summer data separately.

The dataset also includes 5 habitat variables collected for each site (i.e., ba, std, dwd cht, and snd).

I've been instructed to used stepwise regression to complete the analysis for the spring and summer data separately.

I ran a t-test and noticed there are significant year effects in the species data (i.e., spring 2012 vs 2013, and summer 2012 vs. 2013).

My question is, how do I perform a stepwise regression on the grouped spring 2012/2013 data taking into consideration year effects?

Thank you for your assistance - Mark

Trusted Advisor
Posts: 1,436

Re: Stepwise regression with year effects

PROC REG will do a stepwise regression

It's not clear to me that stepwise is appropriate here, as you seem to have described a nicely designed experiment, where the independent variables should be orthogonal to one another. If that is true, you really do have orthogonal independent variables, then stepwise is meaningless waste of time. If I were you, I would fit the model using either PROC GLM or PROC GLIMMIX and ignore the suggestion to do a stepwise regression.

New Contributor
Posts: 2

Re: Stepwise regression with year effects

Hi Paige.  Thank you for the helpful advice.

I'm new to SAS and not familiar with PROC GLIMMIX.  Could you please help me with the syntax?

Much appreciated - cheers Mark

Trusted Advisor
Posts: 1,436

Re: Stepwise regression with year effects

I'm new to SAS and not familiar with PROC GLIMMIX.  Could you please help me with the syntax?

There's a lot of details about your design that you haven't explained clearly, you haven't discussed your response variables, and in any event, I am not an expert on repeated measures designs in this context.

We need a much clearer explanation of the entire design and the response(s).

Grand Advisor
Posts: 16,926

Re: Stepwise regression with year effects

That design doesn't lend itself to stepwise regression.

Ignore that component and use the appropriate analysis method for the design.

Ask a Question
Discussion stats
  • 4 replies
  • 224 views
  • 0 likes
  • 3 in conversation