Contrasts and correlations in theory assessment

Area of Science:

• Quantitative Psychology
• Statistical Modeling
• Research Methodology

Background:

• Assessing competing theories requires robust methods for evaluating their predictive accuracy.
• Traditional effect sizes may not adequately capture nuanced comparisons between multiple conditions or theories.

Purpose of the Study:

• To present a systematic, quantitative approach for evaluating predictions from competing theories.
• To introduce and demonstrate the utility of contrast and correlational indices of effect sizes.

Main Methods:

• Utilized contrast F and t statistics for comparing and combining predictions with continuous data.
• Employed z contrasts for analyzing frequencies in 2 x k contingency tables.
• Demonstrated the calculation of contrast correlations from published results.

Main Results:

• Contrast correlations offer an effective way to estimate effect size correlations, particularly in designs with more than two groups.
• The alerting correlation provides an alternative for assessing a contrast's predictive power.
• Omnibus F-tests are often less informative than focused contrast analyses due to their broad scope.

Conclusions:

• Employing focused questions through contrasts significantly improves the clarity and statistical power of research answers.
• This approach enhances the precision of theoretical comparisons and data interpretation.