Daniel posted on Tuesday, September 02, 2003 - 8:27 am
Hi, I have been asking a lot of questions about Monte Carlo models. The reason is that I am trying to estimate the power to detect moderation in an LGM. Is the Monte Carlo method the best approach for this aim? If so, what is the ideal variance covariance matrix to use. I have data for four waves. We want to project for an additional 4 years. The question of what covariance matrix to use is a complex one for me, given that I already have some data. As usual, I'd appreciate your guidance.
I think Monte Carlo is a good way for your model. Another option is the Satorra Saris approach which can be used with no missing data and no non-normality.
If you don't have theoretical parameter values to use for data generation, then the best you can most likely do is look at your sample data. There could be many criticisms of this. But I don't know of any alternative.