How do you choose starting values for OpenMx models?
you want them to be close but not the exact answer
for a mean you could check the mean of the data and then add a random number about .5 a sd
the optimiser needs to be able to travel to arrive at the right answer
In theory, your starting values should not matter. The optimizer should get to the right answer no matter where it starts. In practice, you want the starting values to be plausibly in the neighborhood of the right answer.
mxTryHard() will wiggle your starting values and try a few different runs.
For some models, mxAutoStart() will set pretty good starting values, but it doesn’t work for all models.