A reproduction of the results of Onyike et al. (2003)

Abstract

Onyike et al. (2003) analyzed data from a large-scale US-American data set, the Third National Health and Nutrition Examination Survey (NHANES-III), and reported an association between obesity and major depression, especially among people with severe obesity. Here, we report the results of a detailed replication of Onyike et al.’s analyses. While we were able to reproduce the majority of these authors’ descriptive statistics, this took a substantial amount of time and effort, and we found several minor errors in the univariate descriptive statistics reported in their Tables 1 and 2. We were able to reproduce most of Onyike et al.’s bivariate findings regarding the relationship between obesity and depression (Tables 3 and 4), albeit with some small discrepancies (e.g., with respect to the magnitudes of standard errors). On the other hand, we were unable to reproduce Table 5, containing Onyike et al.’s findings with respect to the relationship between obesity and depression when controlling for plausible confounding variables—arguably the paper’s most important results—because some of the included predictor variables appear to be either unavailable, or not coded in the way reported by Onyike et al., in the public NHANES-III data sets. We discuss the implications of our findings for the transparency of reporting and the reproducibility of published results.