I am estimating a latent growth model in which the outcomes are repeated administrations of the same tasks. I was wondering if it is possible to estimate practice or re-testing effects as part of the model. This will allow the study of the impact of retesting on outcomes. Is it possible to estimate this in MPLUS? Would it be possible to estimate a retest mean and variance as well as correlations with other latent components? Could you point me to a reference or example using MPLUS? Thank you! Michelle
I am not quite sure what to say here, so let me give you the associations that come to mind. Don't re-test effects have to do with correlations across time? The growth model already does model that. Are you thinking residual correlations for the outcome over time? When you say retest mean, variance, and correlations I am thinking methods factor, but I don't see how a methods factor can be added and be identified unless some type of MTMM measurement structure has been used. I have not seen this discussed in the literature.
Practice effects would seem to be part of the basic idea of what a growth model describes in the sense that practice would result in growth, that is, a slope growth factor with a positive mean. So like a learning curve, possibly not linear. So that would then be just a regular growth model.
Dr. Muthen, Thanks for your response. I was referring to a latent growth model in which retest is included as a latent variable along with intercept and rate of change similar to what is described in McArdle, J.J., Fisher, G.G., & Kadlec, K.M. (2007) Latent Variable Analyses of Age Trends of Cognition in the Health and Retirement Study. Psychology and Aging, Vol. 22, No. 3, 525-545. I am interested in fitting a similar model in MPLUS. The authors in the article used SAS Proc Mixed. Michelle
I think you are referring to their page 541 left-column discussion of adding an “R[t]” term, which affects the growth curve at all time points but the first. For some reason this section seems to have switched from a single-level, wide, multivariate approach to growth to instead use SAS PROC MIXED to do a 2-level, long, univariate approach. This could just as well have been done in Mplus. In Mplus, the R[t] term is a within-level covariate with a random slope specified on the within level and with mean and variance estimates given on the between level.