creating an ESTIMATE statement

[j4sanford]

Using a WEIGHTED dataset, I am running logistic regression and have encountered two interactions: one for sex and one for race among a model comprised of 5 exposure variables and 1 outcome variable. After identifying where the interaction lies, how do I stratify my data considering that it is a weighted data set? I have been told that I need to create a flag variable or an LSMestimate statement. However I am pretty new at SAS coding and need a step by step example of how my coding should look. Also, with me using PROC  SURVEYLOGISTIC, an oddsratio statement does not work. Help please!!

[Reeza]

I know I recommended you post this here, but I also recommend you post this issue on stats.stackexchange.com. Please review their protocol for asking a question before you post, your question, as stated here it will not meet the minimum requirements on their site. 

This question is a bit of statistical methodology as well as coding issues, which is why I'm recommending you post on Cross Validated. 

[j4sanford]

Thanks, I'll give it a shot!

[StatDave]

For most purposes (such as comparisons of levels of a predictor), it is much easier to use the LSMEANS or LSMESTIMATE statement than the ESTIMATE statement. This is particularly true in models that contain interactions because people have considerable trouble properly determining the linear combination of model parameters that should be specified in the ESTIMATE statement. This note illustrates using all of these statements.  Read the Introduction first, then see the examples that use these statements.  What is discussed in this note is general and applies to any modeling procedure, so don't ignore the examples that happen to use PROC GLM.