PROC CALIS computes scoring coefficients which can be used by PROC SCORE to compute estimated factor scores for individual observations. In the PROC CALIS output, these coefficients are called "Factor Score Regression Coefficients." In the OUTSTAT= data set created by PROC CALIS, these coefficients correspond to observations where the _TYPE_ variable takes on the values LSSCORE or SCORE.

The problem is that PROC CALIS computes these coefficients from the unrotated pattern matrix even if a rotation is specifed on the FACTOR statement. The Factor Score Regression Coefficients will be correct only when specifying ROTATE=NONE. Note that by default no rotation is done, so the coefficients are correct if ROTATE= is not specified.

For any other value of ROTATE=, the coefficients will be incorrect. If you use PROC SCORE to create estimated factor scores using the coefficients, the factor scores computed will also be incorrect.

It is not obvious that the values are not correct.

One way to obtain scoring coefficients computed from a rotated pattern matrix is to use PROC FACTOR for the analysis instead of PROC CALIS. Another option is to pass the pattern matrix created by PROC CALIS to PROC FACTOR and allow PROC FACTOR to rotate the pattern matrix. PROC FACTOR will create Standardized Scoring Coefficients which can be used along with PROC SCORE to create the estimated factor scores taking the rotation into account.