|
|
Random-Intercept LTA (RI-LTA) |
|
Message/Author |
|
|
Dear Drs. Muthén, how would the new RI-LTA approach be combined with the "profile similarity testing" procedure from Morin & Litalien (2017) [cited under the LTA topic on the "Papers" page]? Specifically, I am wondering 1. whether the RI would be included as soon as the model contains more than a single time point (i.e., any test of profile similarity)? Or is the RI only relevant with "true" LTA models estimating transition probabilities (e.g., C2 ON C1)? 2. how an RI would be incorporated when manually assigning class probabilities (see Morin & Litalien, 2017, pp. 10-12)? Thanks so much! |
|
|
1. I think RI is relevant only with C2 ON C1 models. 2. I don't know the point of that approach for LTA. Is your interest in testing measurement invariance across time? |
|
|
The point is akin to testing measurement invariance, in that latent class parameters (number of classes; means; variances; class sizes) are successively constrained to equality over time. My interest is in running those tests before estimating transitions, because I want to know how similar those profiles, between which individuals are transitioning, actually are. |
|
Back to top |
|
|