|
|
GMM with different growth parameters |
|
Message/Author |
|
Linqin Ji posted on Wednesday, August 04, 2010 - 8:17 am
|
|
|
The following is my commands for GMM with the same growth parameters for all trajectories. How could it be transfered to commands for GMM with different growth parameters for different trajectories, such as a only intercept model for class 1, and 7, intercept and slope model for class 2 and 4, intercept, slope, and q model for class 3 and 6, and a intercept, slope, q and cu model for class 5. Thanks. DATA: FILE IS 1.dat; VARIABLE: NAMES ARE bh gender y1-y4 y5-y8; usev are y1-y4; missing are all (-999.00); idvariable is bh; CLASSES = c (1); censored = y1-y4 (b); ANALYSIS: TYPE = MIXTURE; STARTS = 20 2; MODEL: %OVERALL% i s q bu| y1@0 y2@1 y3@2 y4@3; OUTPUT: TECH1 TECH8; plot: type = plot3; series = y1-y4(s); |
|
|
You need to specify the model using the class-specific part of the MODEL command, for example, %c#1% [s@0 q@0 bu@0]; s@0 q@0 bu@0; for an intercept only model for class 1. By the way, I'm not sure a cubic model is identified for four timepoints. |
|
Linqin Ji posted on Thursday, August 05, 2010 - 1:48 am
|
|
|
Why we need the command 's@0 q@0 bu@0;' which means to set the variance of s, q, and cu to be 0? |
|
|
The general model is specified in overall. If you want class one to be an intercept only class, then the means, variances, and covariances of the linear, quadratic, and cubic growth factors need to be fixed at zero so that class has only an intercept growth factor. When the variances of the growth factors are fixed at zero, the covariances are fixed at zero automatically by Mplus. |
|
Back to top |
|
|