Laboratory for latent variable modeling
Latent variable modeling
In most psychological and social science research, variables of central interest, such as intelligence, love, happiness, depression, self-esteem, confidence, relationship, etc., are not directly observable. These hidden or latent variables can only be inferred from observed data. We are primarily interested in developing and applying new statistical tools to understand the nature and dynamics of latent variables.
General statistical framework
A variety of methods have been developed to analyze latent variables such as factor analysis, structural equation models, item response theory, multilevel models, latent class models, etc. We understand these different methods in a general statistical framework, which offers us great freedom and flexibility in building, developing, and estimating a variety of models without moving from one analysis to another.
Statistical analysis requires a computational process to estimate parameters based on observed data. We are interested in developing efficient estimation algorithms and software for estimating complex latent variable models.