create a composite index (principal component analysis)

[LanMin]

Dear All,

I will run a regression analysis with several independent variables, e.g., A, B, C. If my hypothesis predicts a composite measure of A,B,C should also be used as an alternative  independent var,

How do I create this composite index using A,B,C. ?  Do I use principal component analysis or other methods?

that is my model may look like

Y = composite (A,B,C)

if you could share some sample code, that will be great!

Lan:smileysilly:

[Reeza]

Usually they hypothesis would specify the composite measure .

If not then yes, PCA could be used, and assuming you use the first component only.  But you should check the loading factor as well to make sure it makes sense.

Factor Analysis would also work (similar concept, but different).

You can look at PROC PRINCOMP as well as:

PROC PRINQUAL
PROC FACTOR