topic Re: What is the difference between the usage of proc factor and princomp for doing PCA analysis? in New SAS User

What is the difference between the usage of proc factor and princomp for doing PCA analysis?

sdhilip — Sat, 20 Apr 2019 06:02:23 GMT

I am doing PCA analysis in SAS Guide. I have a Pizza dataset which has a brand, ID and 7 variables.

Method: 1

I am using proc factor and finding the eigenvalue and interpret the result from the rotated factor pattern. The method used: Principal

proc factor data=work.pizza1
method=principal scree mineigen=0 NFACTORS=7 ROTATE=varimax REORDER OUT=PIZZA1;
var mois--cal;
run;

Result:

I reduced three variables from the rotated factor pattern.

Method: 2

I am using proc princomp.

proc princomp data=work.pizza1 out=new;
var mois--cal;
run;

I am getting eigen value same for both the methods as it is derived from the correlation matrix. However, value is different for rotated factor pattern and eigen vector .

Result:

My question:

1) What is the difference between princomp vs proc factor while using PCA analysis?

2) I am interpreting the result from rotated factor pattern in proc factor. How to interpret the result from princomp? Should I use eigenvector to interpret the result? If so, it shows a different one compared to proc factor.

3) Are the results same if I run both factor and princomp?

4) How to use the method in princomp? Here, I used varimax.

I attached the output and sas code.

Please advise.

Re: What is the difference between the usage of proc factor and princomp for doing PCA analysis?

PaigeMiller — Sat, 20 Apr 2019 11:21:57 GMT

The difference is that you asked PROC FACTOR to rotate the components. The PROC PRINCOMP output does not rotate the components.

Re: What is the difference between the usage of proc factor and princomp for doing PCA analysis?

sdhilip — Sat, 20 Apr 2019 21:40:29 GMT

Thanks @PaigeMiller Which result should I take proc factor are princomp? Which one should I go if I asked to do a PCA analysis? Please advise.

Re: What is the difference between the usage of proc factor and princomp for doing PCA analysis?

PaigeMiller — Sun, 21 Apr 2019 11:27:31 GMT

There is no possible answer to this without a lot more information about the goals of the analysis. Rotated or unrotated components both have usefulness, depending on what you are trying to do with the components.