Ninette posted on Wednesday, January 21, 2015 - 12:18 am
Dear Linda, my longitudinal model includes two timepoints of questioning. To be included in the estimates, persons have to meet requirements. Some persons only meet them at one timepoint, this means I would like to include them maybe at the first, but not the second time point. I tried to solve the problem by setting missings on all variables of the excluded time point, but I am not sure if this is the right way. For my model I am using a Bollen-Stine-Bootstrap with FIML. To exclude such persons wouldn't be the right way I think, because my sample would be an N=300 instead of 530. Thank you!
The treatment of missing depends on how the persons are excluded. Does the exclusion depend on the observed variables in the model, does it depend on other variables? Handling it via FIML assumes that the missingness is either totally at random or a function of the observed variables in the model.