Re: FA vs. PCA and grouped vs. non-grouped

One last thing, before I let this rest in peace! John mentions that
"SAS probably has no method to estimate factor scores when you've got
polychoric correlation". Now there is a macro produced by SAS called
'Polychor' that, according to SAS, "create[s] a matrix of tetrachoric
correlations (for binary variables) or polychoric correlations (for
ordinal variables)". Since PROC FACTOR can use either a dataset or a
correlation matrix, I could then use this procedure to output factor
scores using the matrix of polychoric and tetrachoric correlations.
Since you mention that this is usually a complex task, I'm wondering if
this solution is not too simple to be the right one. One concern I
would have would be to know what this macro produces when it computes
the correlation between an ordinal and a binary data. This latter part
may be more suitable for the SAS-L group but if this macro does what it
says it does, do you believe this is a suitable option?

Thank you,

Daniel

.