Abstract

This dissertation proposes a graphical analysis and presentation system for fitting, evaluating, and reporting longitudinal models in social sciences. The graphical innovations demonstrated here address practical issues that arise in evaluating sequences of statistical models. A progression of nested or otherwise related models in a sequence creates a context for model comparisons. The proposed graphical methods provide the researcher with visualization tools to facilitate model evaluation, using data mapping and interactive document design. The study applies these methods to examine empirical trends of religious involvement using a nationally representative household sample of American youth, the National Longitudinal Survey of Youth, 1997 (NLSY97). Annual measures in the NLSY97 from 2000 to 2011 provided panel data on church attendance from approximately 9,000 individuals born between 1980 and 1984. These data are examined using latent curve models (LCM) to study the nature of change in religious involvement between ages 13 and 31. Data, code, and reproducibility instructions for this study are published as a GitHub repository and are available to the research community.