Methodological Research

Latent Variable Modeling

Structural equation modeling, longitudinal data analysis, and principled treatment of missing data.

Back to Research
Overview

About this work

This line of work concerns latent variable models — structural equation models, latent class and mixture models, and growth models for longitudinal data — and the statistical challenges that arise in estimating them, such as label switching, class enumeration, and missing data.

A recurring theme is understanding when these models recover meaningful structure and how to quantify uncertainty in the conclusions they support.

Selected Work

Related publications

2026

Class selection in growth mixture models: Comparing information criteria to nonparametric and parametric Bayesian approaches

S. Depaoli, M. Qiu, H. Liu, & M. Jauregui
Structural Equation Modeling
2024

Label switching in latent class analysis: Correct assignment of individuals, accuracy of parameter estimates and confidence intervals

M. Qiu & K.-H. Yuan
Structural Equation Modeling