Anonymous posted on Friday, November 04, 2005 - 9:30 am
Hello - I'm a relative beginner with Mplus and very rusty with SEM with some training but little experience running latent models. I was wondering if you could recommend a beginner's guide to simply getting a model to run without errors! E.g., a hierarchy of things to check if a model won't converge, steps to investigate causes of a non-positive definate covariance matrices, etc. In my prelim analyses I'm running a LGM with one dichotomous predictor, about 200 Ss, and the data are structured as a cohort sequential design with a wide age range. There are some small "cell" sizes, e.g., only 5-10 kids with data at youngest age (11) and almost all values are zero at age 11, etc.
If you look under non-convergence in the Mplus User's Guide index, you will find some suggestions about convergence problems. I would suggest builiding up the model stepwise. You can also send your input, data, output, and license number to email@example.com and we can take a look at it.