SAS introduction, part 3

Factor analysis

For this we will use PROC FACTOR.

The following covers the use we will make of PROC FACTOR today:

proc factor data=<sas-data> cov nfactors=<n> rotate=<rot> all;

DATA=<sas-data> <=> input dataset specification.
COV <=> means analyse variance covariance matrix, otherwise analyse the correlationmatrix.
NFACTORS=<n> <=> Number of factors we think are "significant" or "reasonable", e.g.
NFACTORS=2 which are kept for the rotation below.
Nothing means FACTOR retains all factorer with eigenvalues>1.0
ROTATE=<rot> <=> Rotation criterion. ROTATE=NONE or exclusion of "ROTATE" means no rotation.
ROTATE=VARIMAX means rotate using the varimax criterion.
ALL <=> Means we want all possible output, exepting certain plots which must be requested seperatly.

Exercises

1. Perform a factor analysis on all the variables of SUNDHED. An analysis using the correlation matrix suffices.
2. Perform another factor analysis, retaining a "suitable" number of factors. Rotate using the VARIMAX criterion.
3. See if you can interpret the above analyses. A simple way of starting an interpretation is by finding the highest absolute value for each row in the "factor pattern". (In this way any variable will be included in the explanation of one factor only.) Then try to interpret the factors by just looking at those variables.