2.3. Analytic Approach

First, we estimated descriptive statistics using sampling weights. Subsequently, we estimated four multilevel logistic regression models, controlling for country-level clustering, to examine the associations between gender inequality and caregiver physical abuse. Model 1 estimated the odds of physical abuse as a function of country-level GII and MICS round. Model 2 added household characteristics to Model 2. In Model 3, we estimated the odds of physical abuse as a function of GII and child gender, after controlling for caregiver, child, and household characteristics. Model 4 estimated the moderation between GII and child gender. After estimating the multilevel logistic regression analyses with the GII and child gender interaction, we calculated predicted probabilities of abuse to more fully explore these associations. Analyses were performed using Stata SE version 15 [47].