Shige posted on Monday, January 02, 2006 - 7:46 pm
Dear Linda and Bengt,
I have a question about how to determine sample size via Monce Carlo simulation using Mplus.
In Bengt & Bengt (2002), you desmonstrate how to decide sample size via simulation fora group of CFA models and a group of growth models. In the simplest CFA case, the desired sample size is proven to be 150; in a some what more complicated CFA model it is 175, etc. My question is about the general strategy that leads to the above sample size estimates: do we have to try out different number (sample size) in the simulation (I tried 140, 100, 160, for example), check the model output and hand calculate the set of numbers you mentioned in pp.8 (and use the three creterion you mentioned on the same page)?
Also, could you point me to some other resources on this topic? This is an extremely facility for researchers. Thanks!
Yes, you would follow the strategies suggested in the paper. You could also see the 1997 Muthen and Curran paper in Psych Methods.
Thomas Wan posted on Tuesday, August 01, 2006 - 10:46 am
Dear Linda and Bengt: One of my students has a path model with 8 variables. She has a total of 120 subjects. The model has a total of 16 free parameters, 20 degrees of freedom, and 36 sample moments. How can we simulate the model to determine the power of the sample size?
There are several examples of path analysis in Chapter 3. The data for all of these examples was generated using the Mplus Monte Carlo facilities. The Mplus CD has these Monte Carlo inputs on it. I would select the path model most like yours and start with the Monte Carlo example for that example.