You are here

tryHard output

2 posts / 0 new
Last post
mirusem's picture
Joined: 11/19/2018 - 19:32
tryHard output

I was kind of surprised to see (before I used the tryHard option) not having any issues per se, but when I did incorporate it, it looks like sometimes the output of a phenotype would be, for example: Retry limit reached; Best fit=-353.49615 (started at -352.57463) (11 attempt(s): 2 valid, 9 errors)
Does that mean it was a success and it just picks one of the valid solutions? It is certainly different from the output:
Solution found! Final fit=-332.97766 (started at -295.8754) (2 attempt(s): 2 valid, 0 errors).

Just want to be sure there's nothing problematic in this case (also posted this on the umx forum specifically, since that's what is calling tryHard in this case).

I appreciate it!

AdminNeale's picture
Joined: 03/01/2013 - 14:09
Possibly, one valid is ok

It's possible that one valid solution was found. Sometimes code 6 happens a lot - especially with FIML and ordinal/binary data. However, it is possible that your model isn't identified - that will cause mxTryHard (well most any optimization) problems.

One way to gain confidence that the model is identified is to repeat optimization from different starting values, which is what mxTryHard() does. If the "valid" solutions found have the same parameter estimates and the same -2lnL, and this is the lowest -2lnL of all the attempts, it's likely the model is identified. You can check this at the solution by mxCheckIdentification(modelAfterItWasRun) (or whatever your model is called).