Statistical Methodology for the Analysis of Repeated Duration Data in Behavioral Studies

Area of Science:

• Behavioral Science
• Statistical Modeling
• Data Analysis

Background:

• Repeated duration data are common in behavioral studies.
• Classical linear or log-linear mixed models are often insufficient for analyzing non-negative, skew-distributed duration data.

Purpose of the Study:

• To recommend a statistical methodology specifically designed for duration data.
• To address the limitations of traditional models in analyzing repeated duration data.

Main Methods:

• Proposing a methodology based on Cox mixed models implemented in R.
• Utilizing simulations and quantile-quantile plots to compare log-linear and Cox mixed models for goodness-of-fit.

Main Results:

• Demonstrating the limitations of linear and log-linear mixed models with speech and gesture interaction data.
• Showing that Cox models are validated on the data, unlike linear models.
• Identifying a significant effect using Cox models that linear models missed.

Conclusions:

• Providing methods for selecting the best-fitting models for repeated duration data.
• Confirming that Cox models are best suited for analyzing the studied data set.
• Highlighting the superiority of Cox models over linear and log-linear models for duration data analysis.