Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.
From | Dave Garson <garson@ncsu.edu> |
To | statalist@hsphsun2.harvard.edu |
Subject | st: More re factor loadings |
Date | Fri, 27 Sep 2013 14:45:34 -0400 |
Regarding my recent queries about mult. correlation and factor loadings in sem: (1) Multiple correlations: Kristin's recommendation of estat eqgof, which I had overlooked, does generate the same multiple correlations as SAS and SPSS. Thanks. (2) Doing sem, standardized and using the indicator path coefficients does not give the same factor loadings. The Stata SEM manual does indeed refer to these as factor loadings, but this usage of the term differs from SAS and SPSS and the coefficients differ, or so it seems. My question remains whether the factor loadings can be obtained in Stata. To be more specific, this is what SPSS labels "factor score weights" and SAS labels "latent variable scores regression coefficients." I have seen them also called factor weights, factor loadings, factor pattern coefficients, or validity coefficients. In my understanding, the factor loadings that the Stata SEM manual and Kristen reference are the path weights from the latent variables to their indicator variables, based on the observed covariance matrix (or correlation matrix) input. What SAS and SPSS are referring to are the regression weights used when calculating the latent variables from the observed (indicator) variables on the basis of the matrix of covariances implied by the model for all variables in the model. Again, tips on how to get what SPSS and SAS are referring to would be helpful. Best wishes, Dave * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/faqs/resources/statalist-faq/ * http://www.ats.ucla.edu/stat/stata/