mxTryHard and loglikelihood differs from other runs of a model

A user asks
> When we tried to run the tutorial script from your preprint PeerJ paper (Bates, Maes and Neale, 2017), our script doesn't produce the same overall estimate or standardised estimates as in the PeerJ paper (e.g. we have 27,287.23 as the -2*loglikelihood instead of 12,186.28 in the PeerJ paper).

Generally, if you’re in any doubt about a model solution or if there’s a code or warning after the fun, it's worth trying:

    m1 = mxTryHardOrdinal(m1)

It might be too that you are using a different default optimizer use this code to see the options:

umx_set_optimizer()