What do convergence warnings in a hierarchical model mean?

[Wanja Mössing]

Hi everyone,

I've been using the Memtoolbox for a while now and find it to be a very helpful tool. Thanks a lot for providing it!

Recently, I tried to fit a hierarchical model (Standard Mixture with kappa) for 40 subjects. During the computations, I received 4 warnings about convergence (see below). The same happens if I call MAP directly. A colleague of mine gets a similar warning for a similar procedure. We investigated the code and searched online but couldn't really track down what precisely is not converging, and why it is not converging. It'd be great if someone could explain what this means, how serious of a problem this is (how unreliable are the resulting estimates?), and whether there's something we can do about it.

We're using Matlab R2019a with the optimization toolbox installed, in case that's relevant.

Thank you!
- Wanja

MTB will now fit the model to your datasets...
Warning: Maximum likelihood estimation did not converge.
Function evaluation limit exceeded.
> In mlecustom (line 239)
  In mle (line 245)
  In MAP (line 46)
  In MemFit>MemFit_MultipleSubjects (line 368)
  In MemFit (line 138)
  In TM_05_hierarchical_MCMC_standardmixturefit (line 53)