This is the setup of the LCS constant change model that replicates the LGM linear model, first shown by

McArdle, J. J. (2001). A latent difference score approach to longitudinal dynamic structural analyses: a festschrift in honor of Karl Jöreskog. In R. C. K. G. Jöreskog, S. H. C. D. Toit & D. Sörbom (Eds.), Structural equation modeling: Present and future (pp. 342–380). [:353]

Also, there is a simple way to replicate a 2 wave LGM model as a LCS with constant change, the setup below does not look simple because I chose to specify EVERY parameter possible:

## Tuesday, July 30, 2013

## Monday, July 15, 2013

1. Run a paired t-test to answer the question: 'Did outcome Y change from baseline to follow-up" - in AMOS v. 5 [free software, this version only] and in Mplus.

2. Compare the paired t-test results to a flexible model where the change is allowed to

*depend*on the initial values of Y (which the paired t-test does not allow).

3. Compute statistical power for the 2 models, and see which fares best, and why; this can be easily done with very simple syntax ran in Mplus.

For more detailed input and output files to replicate all our analyses, go to:

http://trippcenter.uchc.edu/modeling

The changes seem dramatic when outcomes are not centered [see last image below], less impressive when wave 1 and wave 2 measures are centered on the wave 1 mean [right below], still some differences are notable, see 2nd graph of p values.

Subscribe to:
Posts (Atom)