|
|
MILM with dropouts and WLSMV estimation |
|
Message/Author |
|
|
Dear Dr Muthen, my idea is to fit a growth model for a sample of 574 students on one latent factor repeatedly measured by 13 binary indicators on three occasions as a multiple indicator (linear) growth model (similar to Wu, Liu, Gadermann & Zumbo 2010). I would like to use the WLSMV estimator but I am wondering if I have to take into account the fact, that there are just 74 students with data for all three waves. The other 500 students missed exactly one occasion, whereas it could be assumed that this is MAR. [Students with missings on two time points are excluded from the analysis because they are NMAR.] My question is whether it is necessary to implement something like diggle-kenward-selection-model, pattern-mixture-model or the like? Or is it possible doing the analysis without any further specification due to the MAR assumption? |
|
|
I would recommend using ML which handles MAR, where MAR is good assumption with longitudinal data. You have only 3 dimensions of integration so that should be computationally feasible. |
|
Back to top |
|
|