Alternative Models for Small Samples in Psychological Research: Applying Linear Mixed Effects Models and Generalized...

Area of Science:

• Psychology
• Quantitative Psychology
• Behavioral Science

Background:

• Psychological research often faces challenges with small sample sizes, particularly in studies involving scarce populations or expensive procedures.
• Traditional statistical methods like listwise deletion can significantly reduce data in longitudinal designs, limiting analytical power.
• Existing techniques struggle to model complex correlation structures common in repeated measures data.

Purpose of the Study:

• To explore and provide a tutorial on two alternative longitudinal modeling techniques: linear mixed effects models and generalized estimating equations.
• To demonstrate the application of these methods in analyzing a small, repeated measures dataset.
• To compare the utility of these advanced techniques for handling limited data in psychological research.

Main Methods:

• Application of linear mixed effects models to a repeated measures dataset.
• Application of generalized estimating equations to the same dataset.
• Comparative analysis of results obtained from both modeling techniques.

Main Results:

• Both linear mixed effects models and generalized estimating equations provided comparable results for the primate attachment and social stress study.
• Each modeling technique offered unique insights, highlighting their distinct advantages.
• The study demonstrated the effectiveness of these methods in preserving data integrity with small sample sizes.

Conclusions:

• Linear mixed effects models and generalized estimating equations are valuable alternatives to traditional methods for analyzing limited longitudinal data in psychology.
• Researchers can leverage these techniques to gain robust insights from scarce or costly datasets without compromising data quality.
• The choice between linear mixed effects models and generalized estimating equations depends on the specific research questions and data characteristics, as each offers unique information.