Combining Latent Variable Modeling Techniques with Cross-classified Mixed Effect Models for Use in Longitudinal Research
Project InformationPrincipal Investigator: Leslie Hawley
Funding Agency: UNL ORED
Award Date: March 19, 2015
Project URL: N/A
For more information please contact Leslie Hawley at firstname.lastname@example.org.
One of the most difficult methodological challenges is measuring human growth and contextual aspects, which add an additional degree of complexity. In particular, there is a gap in the literature regarding the ability to model both changes in context over time (e.g., students who change schools) and multivariate outcome measures (e.g., multiple items measuring a single construct).
Ignoring changes in environmental context over the course of a longitudinal study can lead to an underestimation of standard errors. Additionally, measuring a construct with a single summary score as opposed to using multivariate measures can lead to increased measurement error. The combination of these two aspects is important for longitudinal research; these situations increase the degree of model error, which decreases the power to detect significant effects.
The objective of the proposed project is to evaluate the efficacy of blending a particular type of hierarchical linear model with multivariate latent variable modeling techniques. The proposed plan is to evaluate two types of modeling approaches (i.e., models that either ignore or account for changes in context) across a series of univariate and multivariate measurement model structures.
Data from three longitudinal school-based research studies will be used to investigate the effectiveness of the proposed models. Empirical findings will be used to inform a large-scale Monte Carlo study, which will be submitted for National Science Foundation funding. The proposed study is essential for the principal investigator to secure external funding for future projects examining the efficacy of blending these statistical approaches.