/f-moderation-sem

A SEM cross-lagged moderation model with early-childhood data

Primary LanguageR

Information

  • The following models examines a longitudinal moderation cross-lagged latent panel model.
  • Data are drawn from a large national sample of data for preschool children. This is a now defunct anlaysis.

File Organization

  • All the files for invariance testing are located in the 'invariance' subfolder
    • Important. The minimal requirement of weak invariance is needed for these models.
  • The 'weighted' folder contains all phases of model testing, specifying the complex survey design in all models.
  • The 'no-weights' folder contains all of the models in the weights folder without specifying the complex survey weights.
  • The 'bayes' folder contains all models to explore the credibility interval, and to explore differences in the speed of model convergence.
    • Note. These models will not allow for complex sample weights.

Confirming Parameter Estimates

MPlus tended to yield correct estimates with an error pertaining to the ratio of parameter estimates to clusters in the multilevel data.

To explore the accuracy of these estimates two additional steps were taken:

  1. The models were estimated without weights to see if the parameter estimates were comparable.
  2. The models were estimated with Bayes estimation techniques in order to see the 95% probability that the estimate fell within a range of estimates, and that this was comparable to the other 2 models.