topic Multiple regression, partial F test with variables added in order in Statistical Procedures

Multiple regression, partial F test with variables added in order

code_blooded — Thu, 14 Nov 2019 20:29:58 GMT

Hi,

I am trying to conduct Partial F test on groups of variables added in order to a MLR.

Say my inital model has 3 variables, I add another 4 variables for my second model and so on..

I need to get the r-squared the delta r and the significance of delta F from one model to another.

I have multiple outcomes some are continuous and some are categorical; same is the case with my covariates of interest

Can anyone suggest what will be a good SAS procedure to do it. I tried using proc glm for this but not sure what option to specify to get the accurate significance of delta F from one model to another.

Thanks!

Re: Multiple regression, partial F test with variables added in order

PGStats — Thu, 14 Nov 2019 21:02:07 GMT

Use proc GLMSELECT to do variable selection (method=FORWARD) and COLLECTION effect(s) to group independent variables for model contribution.

https://documentation.sas.com/?docsetId=statug&docsetVersion=14.3&docsetTarget=statug_glmselect_syntax05.htm&locale=en

Re: Multiple regression, partial F test with variables added in order

PaigeMiller — Thu, 14 Nov 2019 21:24:36 GMT

It sounds like you are looking to compute the Extra Sum of Squares for these models, as described here: http://www.jerrydallal.com/LHSP/extra.htm

If so, PROC GLM is a very good tool to use, you can output any statistic computed by PROC GLM to a SAS data set and then use these statistics in your calculations.