RMKirkpatrick

@tbates I do kind of like the idea of a `doRefModels` argument to summary.MxModel. Anyhow, out of necessity, this is a function I wrote and used today: ``` GenomicSEMFitStats

All of those things are caught at runtime (see below), although the non-character labels should raise a more helpful error message. ``` library(OpenMx) # Off-diag input m

Multilevel regression modeling with ordinal variables (or any other kind of variable with which the identity link isn't used) could be accomplished either by integrating out the random effects by...

> Agreed. Would pretending that the data are continuous and computing MLR (robust) SEs be a slightly-better-than-nothing stop-gap? Perhaps. But `imxRobustSE()` does not handle multilevel data correctly. Its man page...

> Yeah, I figured as much... how effortful to revise, compared to your option 1? Trying to revise `imxRobustSE()` to work properly with multilevel data would be a waste of...

Unrelated to redundant constraints...take a look at the CI details for the model when run with NPSOL, versus when run with CSOLNP after the redundant constraints have been removed. NPSOL...

[180816--dependentconstraints.R.txt](https://github.com/OpenMx/OpenMx/files/2295159/180816--dependentconstraints.R.txt) Consider also the attached script. It uses two equality constraints, where one of the constraint functions is a linear function of the other. That's not something SLSQP can handle...

@tbates 1. I agree it would be wise to stop writing scripts that include the same object more times than necessary. But we can also make OpenMx a bit smarter...

I put some thought into this issue this afternoon. Nelder-Mead just needs to be taught to ask for the rank of the equality-constraint Jacobian at the start values, and warn...

@mhunter1 So, this issue is actually a matter of 3 related but distinct issues. Number 1 is the "counting problem at the end"; we want OpenMx to correctly adjust the...