Tim Bates

This would be pleasant. Nice to keep things in non-scientific where possible (maybe set a threshold of e+-6?), and marking thousands helps parsing: ``` [hh:mm:ss] MxComputeGradientDescent(SLSQP): 40eval/sec LL = 318,876...

`umxFitIndices` has this metric among others. Looks pretty simple: ```R SRMR = sqrt(mean((residual.cor^2)[lower.tri(residual.cor, diag = TRUE)])) ```

Decision was to implement SRMR for WLS in the first instance? See also #198 re AIC

Does WLS now have a comparable fit index? I think the iron gal impetus those years ago was that WLS modelling in openmx didn't offer a suitable fit index to...

FYI, this error for values, would be better if it told the user how many values were expected: ```R mxMatrix(type= "Stand", 5,5, values=1:5) Error: illegal number of elements (5) for...

re using a var name as a latent name: Is that just the special case of when the user declares the variable a definition variable, then that name becomes available...

@RMKirkpatrick perhaps we could start moving files into `inst/model/demonstration` if they meet some criterion for commenting and explanation?

@jpritikin yes - `data` already contains a lot of redundant files though (simple variations on a theme, and no documentation) So we should probably also see whether there are files...

@RMKirkpatrick idea would be to create a folder like passing, still in the test suite, which contains files that are well-enough documented to be useful templates for users.

Sounds good: is `demo` checked in the test suite? If, yes, then let's each of us move (or, more likely copy cut down user-focussed instead of testing-focussed versions) files there...