Flag models with parameter count mismatches within R

[mburnett]

Hello,

I'm building a set of models and returning their results with collect.models. I then use a function with a for loop to extract some of the derived parameters from those models and append them to the results table.

As part of that for loop, I populate a column with the length of results$singular from every model so I can quickly see which ones may be having structural problems. However, I've just realized that some of my models mention singular beta(s) in the full Mark outputs, but results$singular from those models just displays NULL, so I'm overlooking them when I examine my full results table. All of these models seem to feature the following warning:

* * WARNING * * Number of parameters estimated from gap method (=6) and threshold method (=7) differ. Inspect full output.

My question is whether or not there is a way to "catch" these models with mismatched parameter estimates in R so I can flag them in my results table. I'd prefer not to have to look through all of the OUT files to hunt them down, but I cannot find any trace of the warning in the list produced by running mark. In addition, why does results$singular not include the betas mentioned in the full output as being singular for these models?

Many thanks!
Mike

[jlaake]

Mike

Please send me one of the output files where it did not pickup on the singular parameters. RMark function extract.mark.output looks through output file to read results into mark object. I'm guessing that it is not catching it because of differences in wording of output file. The base name of the .out file is included in the mark object so you could use that to read in the output file and find those models that were not found by using grep to search file if you don't want to wait for me to fix it.