The regression trap: why regression analyses are not suitable for selecting determinants to target in behavior change...

Area of Science:

• Behavioral Science
• Health Psychology
• Intervention Science

Background:

• Regression analyses are frequently employed to identify key determinants for behavior change interventions.
• However, the common application of these statistical methods may lead to suboptimal intervention design.

Purpose of the Study:

• To elucidate the limitations of regression analyses in selecting determinants for behavior change interventions.
• To introduce the concept of the "regression trap" and its implications.

Main Methods:

• Theoretical rationale examining determinant overlap.
• Mathematical rationale analyzing regression coefficient formulas.
• Empirical examples using real-world data.

Main Results:

• Regression coefficients can be misleading due to inter-determinant correlations, often neglecting psychological complexities.
• This distortion leads to interventions targeting less impactful determinants.
• Consequently, behavior change interventions may have reduced effectiveness.

Conclusions:

• Theoretical, mathematical, and practical issues render regression analyses unsuitable for selecting intervention targets.
• Intervention developers should simultaneously consider univariate and bivariate data.
• Accessible tools exist to support more appropriate selection methods.