Integrating ANOVA and logistic regression for treatment comparison: A methodological illustration with simulated data

Abstract

Anxiety disorders impair daily functioning and well-being. CBT and MBCT are established treatments whose comparative effectiveness is often examined through a single analytic lens. This didactic study uses simulated data (N = 60; CBT n = 30; MBCT n = 30) to illustrate complementary statistical approaches to between-group differences in post-treatment anxiety, assessed by the Beck Anxiety Inventory (BAI). One-way ANOVA indicated a strong group effect, F(1, 58) = 24.8, p = 6.07×10⠻⠶, Cohen’s d = 1.29, with MBCT showing lower anxiety (M = 18.5, SD = 4.9) than CBT (M = 25.0, SD = 5.2). Binary logistic regression showed each one-point BAI increase was associated with 23% lower odds of MBCT membership (OR = 0.77), with good discrimination (AUC = 0.819) and 76.7% accuracy. These analyses demonstrate that significance testing, effect-size estimation, and classification performance answer complementary questions. Data are fully simulated; findings are not intended for clinical generalization.