10-16-2015 02:36 PM
Anyone familiar with Proc Factor will recognize the REORDER option in the Proc Factor statement. Basically, this option sorts the input variables by the strength of their absolute loading with the latent dimensions, simplifying interpretation of the factor structure. My question is: how is this done? It's not just a simple descending sort by the latent dimensions...there's more to it than that.
Can anybody explain how this option works so that it can be replicated for matrics other than Proc Factor eigenvectors?