|
|
Generate multi-group data within clus... |
|
Message/Author |
|
Dex posted on Saturday, March 24, 2012 - 9:45 pm
|
|
|
Dear Dr. Muthen, I am trying to generate multi-group multilevel data using MONTECARLO command. By specifing "NGROUPS", "NCSIZES", and "CSIZES"; the program generated the grouping variable between clusters by default setting. However, I'd like to generate multi-group data "within" clusters(e.g. students nested within schools, the group vaiable is gender). I was wondering under MONTECARLO, is that possible to do a "personalized" grouping procedure within clusters. Thank you for your time. |
|
|
You could generate a within-level dummy variable on within using the CUTPOINT option. |
|
Dex posted on Sunday, March 25, 2012 - 4:59 pm
|
|
|
Dear Dr. Muthen, Thank you for your reply.In my case, instead of generating a within-level dummy variable as covariate, I hope to specify the group membership(within clusters) independent of all other variables, and I also tried to specify different model parameters for different groups under "model population". Based on my understanding, if using "CUTPOINT" option, the software has to generate the whole thing simultaneously; and I cannot do further constrain on population parameters based on group membership. Could you please give me some further information/help? Thank you and my best. |
|
|
I don't quite understand: You can affect group differences using within-level covariates - the factor means and the factor indicator means can be different across groups. There is a more advanced approach which uses mixture modeling, where within-level latent classes capture group differences. |
|
Back to top |
|
|