Multilevel models for the experimental psychologist: foundations and illustrative examples

Area of Science:

• Psychology
• Statistics
• Experimental Design

Background:

• Multilevel models are widely used in educational and developmental research.
• Their application in experimental designs, especially with repeated measures, is less common.
• Existing methods may not adequately handle complex data structures in experimental settings.

Purpose of the Study:

• To illustrate the utility of multilevel models in analyzing data from experimental designs.
• To demonstrate advanced techniques for handling missing data and modeling multiple outcomes.
• To examine speed-accuracy trade-offs in experimental conditions and at the individual level.

Main Methods:

• Application of multilevel models to two experimental design examples with repeated measurements.
• Techniques for examining categorical, continuous, and semicontinuous independent variables with missing data.
• Simultaneous modeling of response times and error rates using a multivariate multilevel model.

Main Results:

• Multilevel models effectively analyze experimental data with repeated measures, even without a time component.
• The proposed methods properly examine independent variables and handle missing data.
• Simultaneous modeling revealed speed-accuracy trade-offs at both experimental and individual levels.

Conclusions:

• Multilevel models are a powerful and underutilized tool for analyzing experimental data.
• These models facilitate robust analysis of complex experimental designs, including missing data and multiple outcomes.
• The findings support the broader adoption of multilevel modeling in experimental research for deeper insights.