Advised missing data method

Longitudinal analysis > Missings in outcome > MCAR > <75% item missings

In this situation it is advised to analyze the data with a longitudinal model, for example a latent growth model. The longitudinal analysis method that is applied should use some form of full information likelihood estimation, to estimate the model parameters. Only in these kind of models, all the information in the model (also the incomplete information) is used to obtain the most likely parameter estimates for your data. When the missing total scores, are caused by missing item scores, it is important to incorporate the available item information. Since, the item scores are partly observed for the subjects with missing data, some information is available. The item information can be incorporated in the model as auxiliary variables. A manual can be obtained via this link on how to apply such a model and include the item information as auxiliary variables.