Programming the statistical procedures from SAS

Multiple Classification Analysis/Factorial ANOVA in SAS

Reply
Occasional Contributor
Posts: 5

Multiple Classification Analysis/Factorial ANOVA in SAS

Hello,

Can someone help me with how to go about Factorial ANOVA in SAS? It seems this technique is also called as Multiple Classification Analysis.

Thanks a lot!

Respected Advisor
Posts: 2,655

Re: Multiple Classification Analysis/Factorial ANOVA in SAS

This is a pretty general question.  I would recommend reading Introduction to Analysis of Variance Procedures and Introduction to Mixed Modeling Procedures in the SAS/STAT documentation.  There are more than two dozen different procedures in SAS that can do this kind of analysis.

Steve Denham

Occasional Contributor
Posts: 5

Re: Multiple Classification Analysis/Factorial ANOVA in SAS

Thanks Steve for your reply!

Actually I need to find out how many independent variables can SAS take at a time while running MCA in SAS.

I know the syntax used in SPSS which is

anova dept variable by inpdt variable1 (1,3) inpdt variable2 (1,3)/method=hierarchical/statistics=all. 

SPSS can take only 10 independent variables.

I want to know how many can SAS take?

I will go through the document you suggested in order to gain more information. Many thanks

Respected Advisor
Posts: 2,655

Re: Multiple Classification Analysis/Factorial ANOVA in SAS

Since I am not up to speed on SPSS syntax, I don't know if what I am going to say makes sense.  Are you implying that you want a factorial design with more than 10 variables?  For a full factorial with exactly 10 variables, you are looking at something like a 3628800 x 3628800 matrix (approximately), most of which are higher order interactions that are indistinguishable from noise.  On the other hand, if you had 100 variables, and were only interested in second order interactions, the matrix would be 5051 by 5051 (or something close to that), which is probably something that can be fit without too much difficulty in SAS.  You would still need a pretty large dataset to get reasonable interval estimates.

I think to give a more supportive answer, we need more information on the design and possible outcome variables.

Steve Denham

Message was edited by: Steve Denham

Ask a Question
Discussion stats
  • 3 replies
  • 476 views
  • 0 likes
  • 2 in conversation