Restricted Maximum Likelihood Estimation for Parameters of the Social Relations Model

Area of Science:

• Social psychology
• Quantitative psychology
• Behavioral sciences

Background:

• Round-robin designs are common for studying interpersonal judgments and behaviors.
• Traditional analysis of Social Relations Model (SRM) data relies on ANOVA or multilevel methods.
• Existing methods may have limitations in parameter estimation and covariate analysis.

Purpose of the Study:

• To embed the Social Relations Model (SRM) within the linear mixed model (LMM) framework.
• To demonstrate the application of restricted maximum likelihood (REML) for SRM parameter estimation.
• To explore the estimation of covariate effects on SRM-specific effects.

Main Methods:

• The study integrates the SRM into the linear mixed model (LMM) framework.
• Restricted maximum likelihood (REML) is employed for parameter estimation.
• An illustrative example and a simulation study are used for validation.

Main Results:

• The proposed LMM-REML approach provides a unified framework for SRM analysis.
• The method allows for the estimation of covariate effects on SRM parameters.
• Simulation results indicate the proposed approach performs comparably to or better than ANOVA methods.

Conclusions:

• Linear mixed models offer a flexible and powerful framework for Social Relations Model (SRM) analysis.
• The REML estimation method provides reliable parameter and covariate effect estimates.
• This approach enhances the analysis of interpersonal judgment and behavior data.