Two-Step Multilevel Latent Class Analysis in the Presence of Measurement Non-Equivalence

Area of Science:

• Statistics
• Psychometrics
• Social Sciences

Background:

• Latent class models are used for analyzing categorical data.
• Clustered data presents unique statistical challenges.
• Measurement non-equivalence complicates model interpretation.

Purpose of the Study:

• To develop an improved estimation method for two-level latent class models.
• To address non-equivalence of measurement in structural models.
• To provide accurate estimation of latent class coefficients given covariates.

Main Methods:

• Proposed a novel two-step estimation procedure.
• Extended existing two-step methods to incorporate measurement non-equivalence.
• Specified a first-step model to account for covariate effects on measurement.

Main Results:

• The proposed method correctly accounts for measurement non-equivalence.
• Simulation studies demonstrated the properties of the new estimators.
• An applied example illustrated the practical utility of the method.

Conclusions:

• The novel two-step method offers a robust approach for latent class analysis with non-equivalent measurement.
• This technique enhances the analysis of clustered data in the presence of covariates.
• The findings have implications for statistical modeling in various research fields.