I’m asking probably because I’m not totally familiar with missing data handling in SEM. May I ask how is the missing data handled in general by onyx? Is it by imputation? If so, may I ask which type of imputation? Thank you so much!
Onyx handles missing data by using Full Information Maximum Likelihood. When you have missingness in your data which is independent of the actual values or independent after controlling for the available measures (in the literature, for odd reasons, these two cases are called MCAR = Missing Completely at Random and MAR = Missing at Random, respectively), then the FIML estimate is unbiased, and all you loose is the power from the missing values. Imputation methods (with some exception for multiple imputation) will also have no bias for MCAR, but are biased for MAR cases. If you are interested, you can get the point estimates for likelihood based imputation from Onyx by clicking the model and selecting “Estimation” -> “Obtain Latent / Missing Scores”; this will create a new dataset which contains your original data set, but all missing values will be imputed by the maximum likelihood best guess for this value (and, additionally, all latent variables will be contained with their respective scores for all participants).