Latent Model with Longitudinal Health Data

Longitudinal data describes the data that tracks the same type of information on the same subjects at multiple time points. Longitudinal data allows researchers to explore dynamic information rather than static concepts, and it is essential for researchers to understand how participants move from one situation to another by some potential predictors and the association with later outcomes. Variable-centered approach and person-centered approach were two approaches that were often used for longitudinal data analyses. The variable-centered approaches are the traditional approach used to explain relationships between variables of interest in a population. Variable-centered approaches include but are not limited to correlation analysis, regression analysis, latent growth curve analysis, and longitudinal mediation analysis. However, variable-centered approaches are too parsimonious and only provide little specificity across the sampled population. Therefore, the popularity of person-centered approaches has increased rapidly since person-centered approaches provide more nuanced results by categorizing the whole population into different subpopulations. The person-centered approaches include but are not limited to latent transition analysis, latent class analysis, latent profile analysis, and cluster analysis. People’s attention on physical and psychological issues for populations with different ages and cultural had increased rapidly; thus, the analyses that use to identify the relationships between variables of interest over time become essential for promotion of psychological and physical health. For example, demographic and lifestyle variables should be relatively important for population health, but the way to choose the best model that explores these associations should be carefully considered and verified. In addition, there are also many models that can be used to examine the relationships between eating behaviors and potential resulting outcomes (e.g., sleep quality, physical health, cognitive functions), so understand the fit statistics that can be used to select the best model is also essential. Therefore, Chapter I serves as a general introduction to briefly introduce all models and techniques that would be used for longitudinal data analysis. Chapter II has described the bidirectional relationships between night eating, loss of control eating, and sleep quality in Chinese adolescents by latent growth curve analysis and longitudinal mediation analysis. Chapter III has examined the negative emotional eating patterns and associations with demographic information (e.g., age, gender, and BMI-z) and resulting outcomes (e.g., eating disorders, psychological health, and inflexible eating) among Chinese adolescents via latent class analysis, latent class analysis with covariates, latent class analysis with distal outcomes, and latent transition analysis. Chapter IV has explored the latent dietary patterns and associations with cognitive functions and psychological and physical well-being in Chinese older adults by latent profile analysis, multinomial logistic regression, and latent transition analysis. Moreover, there are other analyses/methods that were used in these studies (Chapter II - Chapter IV) such as correlation analysis, bivariate latent growth curve analysis, multivariate latent growth curve analysis, and growth mixture modeling.

Embargo status: Restricted until 09/2028. To request the author grant access, click on the PDF link to the left.