Mon, 11/30/2009 - 22:31

is there internal documentation on the effects of mxOptions and resultant quantities in the @output slots of an MxModel object?

specific problem involves the elements of ...@output$hessianCholesky and ...@output$hessian. for the simple problem i outlined in "summary(mxRun(myMxModel)) gives strange results" i get identical results using all mxOptions defaults and using

this <- mxOption(this, "Hessian", "No"). that is, both ...@output$hessianCholesky and ...@output$hessian give identical results.

there could be a very good reason for this, but on the other hand, it could reflect a bug/documentation error. specifically, does ...@output$hessianCholesky contain a Cholesky factorization of a numerically-differentiated estimate of the Hessian or does it contain the most recent update from, say, a BFGS algorithm? similarly, is ...@output$hessian the estimate from updated Cholesky factors or is it the result of a direct call to numerically estimate the Hessian?

It should be whatever is returned by NPSOL. Neither direct calling nor use of other algorithms, nor algebraic computation is used to compute the Hessian.

I think the NPSOL output presently ends up in a fort.7 file; this is under reconstruction.