Related Experiment Video
Updated: Jun 24, 2025
10:10
Three Differential Expression Analysis Methods for RNA Sequencing: limma, EdgeR, DESeq2
Published on: September 18, 2021
37.2K
Treatments for undefined log ratios in matching analyses.
1Université TÉLUQ, Montréal, Canada.
Journal of the Experimental Analysis of Behavior
|June 5, 2024
Summary
Handling undefined log ratios in matching analyses is crucial. Full information maximum likelihood and omitting undefined ratios offer the best performance for accurate estimations, unlike other methods.
Area of Science:
- Behavioral economics
- Quantitative psychology
Background:
- Matching analyses are vital for understanding behavior-reinforcer relationships.
- Undefined log ratios, occurring when rates are zero, pose a significant challenge, rendering data unsuitable for analysis.
Purpose of the Study:
- To evaluate the effectiveness of five different methods for handling undefined log ratios in matching analyses.
- To compare these methods based on estimation accuracy, sensitivity, and bias through simulations.
Main Methods:
- Simulations were conducted to compare five treatments for undefined log ratios.
- Treatments included omitting ratios, using full information maximum likelihood (FIML), and various constant/mean imputation methods.
Main Results:
- Full information maximum likelihood (FIML) and omitting undefined ratios demonstrated superior performance.
- These methods yielded negligibly biased and more accurate estimates compared to imputation techniques.
Conclusions:
- The study recommends FIML as the preferred method for addressing undefined log ratios in matching analyses.
- Methods involving mean/constant imputation (mean/100, 1/10, +.50) should be avoided due to their poorer performance.
