I would like to find out if it is currently possible to estimate a parallel processes growth mixture model with more than two parallel processes. Moreover, is it possible to estimate a sequential parallel processes model, where, for example, one would estimate the association between two parallel processes in childhood, and their association with two other parallel processes in young adulthood?
I am interested in using these models to study the interconnection between the developmental trajectories of four comorbid psychiatric disorders.
Hello, I run the latent growth mixture models for two parallel processes (offending and victimization) separately for boys and girls. I have identified four distinct classes for boys and three classes for girls. There are two groups of boys with sharply falling offending, one with low level of declining in offending, and one with sharply rising offending. The findings concerning the other construct indicate that boys in classes 1, 3, and 4 demonstrate the declining pattern in the average victimization scores; whereas, boys in class 2 show the increasing pattern of change in average victimization scores over time.
Next, I explore the issue of the developmental linkage between offending and victimization across four latent classes of boys. In order to do that, I calculated the covariance matrix for the latent growth factors for these two domains.
I have got the significant positive slope-slope correlations for each latent class of boys (even for the class in which boys demonstrate the declining pattern in offending but increasing pattern in victimization). I am puzzled by these results?
My question is how to interpret the positive correlation between a negative slope and a positive slope?
Dear Dr. Muthen, I used TECHNICAL 4 OUTPUT, however, I did not pay attention to the warning in the beginning of the overall output which says "WARNING: THE BEST LOGLIKELIHOOD VALUE WAS NOT REPLICATED.THE SOLUTION MAY NOT BE TRUSTWORTHY DUE TO LOCAL MAXIMA. INCREASE THENUMBER OF RANDOM STARTS.
THE STANDARD ERRORS OF THE MODEL PARAMETER ESTIMATES MAY NOT BE TRUSTWORTHY FOR SOME PARAMETERS DUE TO A NON-POSITIVE DEFINITE FIRST-ORDER DERIVATIVE PRODUCT MATRIX.THIS MAY BE DUE TO THE STARTING VALUES BUT MAY ALSO BE AN INDICATION OF MODEL INIDENTIFICATION.THECONDITION NUMBER IS 0.150D-10.PROBLEM INVOLVING PARAMETER 46"
May be that is why I have positive correlation between two slopes of different signs. Arina.