Programming the statistical procedures from SAS

How does SAS/LAB check for confounding in multiple regression?

Accepted Solution Solved
Reply
Occasional Contributor
Posts: 11
Accepted Solution

How does SAS/LAB check for confounding in multiple regression?

Hello,

I'm using SAS 9.2 TS Level 2M2 on Windows XP and notice that under Solutions -->Analysis-->Guided Data Analysis (SAS/LAB), when analyzing the assumptions for mutltiple regression, there is an option to check for "confounded effects."  Can anyone tell me how SAS is checking for confounding effects, or direct me towards any references/documentation?  I can only find very minimal documentation for SAS/LAB and absolutely nothing on what procedures are being used. It is very "black box."

Thanks for any help.



Accepted Solutions
Solution
‎03-10-2013 11:55 AM
Regular Contributor
Posts: 152

Re: How does SAS/LAB check for confounding in multiple regression?

Probably SAS/LAB uses the same rules as those in SAS's procedures for the design of experiments like PROC FACTEX or its ADX system.  Look at the SAS documentation for these procedures (for example, the section on PROC FACTEX related to "Resolution") or the articles at Lex Jansen's Internet site, http://www.lexjansen.com/.

View solution in original post


All Replies
Regular Contributor
Posts: 152

Re: How does SAS/LAB check for confounding in multiple regression?

In SAS/LAB, the meaning of the terms, "confounded effects", differs from that for the terms, "confounding effects".  "Confounded"

implies that some variables are linear combinations of other variables (collinear with these other variables) so that the regression coefficients of the former variables cannot be estimated if the latter variables are present.  I believe the term, "confounded", comes from the literature relating to the design of experiments.  You can view how the former variables are related to the latter variables in an "aliasing structure" from the pull-down menu in SAS/LAB called "Locals".

The term, "confounded", is distinct from the term, "confounding", where a variable may be associated with both another variable of interest and the outcome; adjusting for the confounding variable would change the magnitude and perhaps even the direction of the association between the variable of interest and the outcome.

Occasional Contributor
Posts: 11

Re: How does SAS/LAB check for confounding in multiple regression?

Thank you 1zmm, that was helpful information. I was indeed thinking of "confounding effects."

Still ... does anyone know the procedure/rules SAS/LAB is using to determine if variables are "confounded"?

Solution
‎03-10-2013 11:55 AM
Regular Contributor
Posts: 152

Re: How does SAS/LAB check for confounding in multiple regression?

Probably SAS/LAB uses the same rules as those in SAS's procedures for the design of experiments like PROC FACTEX or its ADX system.  Look at the SAS documentation for these procedures (for example, the section on PROC FACTEX related to "Resolution") or the articles at Lex Jansen's Internet site, http://www.lexjansen.com/.

Occasional Contributor
Posts: 11

Re: How does SAS/LAB check for confounding in multiple regression?

Thanks 1zmm, I will check out those references.  Although I don't know if you're answer is right, it sounds good to me, so I'm marking this as "answered."

☑ This topic is SOLVED.

Need further help from the community? Please ask a new question.

Discussion stats
  • 4 replies
  • 1127 views
  • 3 likes
  • 2 in conversation