help for ^gpfobl^
by Herve M. CACI, March 2003

Rotation after Exploratory factor analysis / Principal components analysis --------------------------------------------------------------------------

^gpfobl^ <matname> [, <rotation(#)> Factors(#)]

Description -----------

This is the translation into Stata programming language of the programs that appear in the paper:

Gradient Projection Algorithms for Arbitrary Rotation Criteria in Factor Analysis

by Coen A. Bernaards & Robert I. Jennrich (http://www.stat.ucla.edu/research)

If this command is used after ^factor^, you have to get the matrix of unrotated loadings. That is,

.matrix A=get(Ld)

.gpfobl A, oblimin(.50)

By default the number of factors is the number of columns of the inputted matrix. However, the user can choose to rotate a lower number of factors.

.matrix A=get(Ld)

.oblimin A,f(3)

Available oblique rotation criteria are:

1. OBLIMIN(gamma): by default, gamma==0.50. If the user specifies gamma==0 then the rotation is QUARTIMIN, and if the user specifies gamma==1 then the rotation is COVARIMIN.

2. OBLIMAX: no parameter needed.

3. IPS: Bentler's Invariant Pattern Simplicity (no parameter needed).

4. CFerguson(kappa): this is in fact a family of rotations. - QUARTIMAX if kappa==0 - VARIMAX if kappa==1/p - EQUAMAX if kappa==k/(2*p) - PARSIMAX if kappa==(k-1)/(p+k-2) - Factor parsimony if kappa==1 where k is the number of the number of factors and p is the number of variables

5. GEOMIN(epsilon): by default epsilon==0.01.