Message/Author |
|
Michael posted on Monday, November 28, 2005 - 6:19 pm
|
|
|
I’m running a GMM model with 3 time points similar to Example 8.6 of the User’s Guide. I’m having difficulty in specifying the correct model command to achieve the relation to the binary distal outcome (DisOut). The User’s Guide indicates that “…since automatic starting values are used, it is not necessary to include class-specific statements in the model command.” (pg 163). I’m specifying starting values for a 3 class solution and have a time varying covariate (TVC) at each measure. I also have antecedent variables (time invariant, x) predicting class membership. I’m unsure as to what the model command would be for the distal outcome (DisOut). %Overall% i s | y1@0 y2@1 y3@2; i s ON x; c#1-c#2 ON x; y1 ON TVC1; y2 ON TVC2; y3 ON TVC3; %c#1% [i*3 s*5]; y1 ON TVC1; y2 ON TVC2; y3 ON TVC3; %c#2% [i*1 s*.5]; y1 ON TVC1; y2 ON TVC2; y3 ON TVC3; %c#3% [i*1 s*-1]; y1 ON TVC1; y2 ON TVC2; y3 ON TVC3; |
|
|
You refer to the distal outcome by its threshold within the %c#% statements. By default these are different across classes which implies that c influences the distal. |
|
|
Hi, I'm a new user and have been going through your video courses. In your LSAY example with high school dropout as a distal outcome, could you tell me how you computed the probability of outcome (e.g. 69% in class 1, the low math achievers) given the output? I'm not sure which part of the output informs us about this; the only thing I see relating the distal to the classes is the threshold. Many thanks, Tracie |
|
|
You get thresholds for each class and the thresholds can be translated into probabilities. See slide 131 of Topic 6 - the alcohol example shows how to compute the probabilities from the thresholds. |
|
|
Thank you, that's very helpful! Tracie |
|
Back to top |