Choosing Imputation Models

Imputing missing values is an important preprocessing step in data analysis, but the literature offers little guidance on how to choose between imputation models. This letter suggests adopting the imputation model that generates a density of imputed values most similar to those of the observed values for an incomplete variable after balancing all other covariates. We recommend stable balancing weights as a practical approach to balance covariates whose distribution is expected to differ if the values are not missing completely at random. After balancing, discrepancy statistics can be used to compare the density of imputed and observed values. We illustrate the application of the suggested approach using simulated and real-world survey data from the American National Election Study, comparing popular imputation approaches including random forests, hot-deck, predictive mean matching, and multivariate normal imputation. An R package implementing the suggested approach accompanies this letter.

Early Infant Feeding of Formula or Solid Foods and Risk of Childhood Overweight or Obesity in a Socioeconomically Disadvantaged Region of Australia: A Longitudinal Cohort Analysis

In southwestern Sydney the timing of introduction of formula and solids may be associated with risk of childhood overweight or obesity, and this may vary by age at breastfeeding cessation during first year. We included 346 infants from southwestern Sydney using the longitudinal study for Australian children (LSAC), who at baseline were singleton, full term, and normal weight births. The outcome risk of overweight or obesity was measured at every two-year interval of children aged 0 or 1 year at baseline until they reached age 10 or 11, defined by body mass index (BMI) ≥ 85th percentile, using the Centre for Disease Control and Prevention growth charts. Age at introduction to formula or solids was dichotomized at four months. We used mixed effects logistic regression for performing all analyses with and without adjusting for mother’s BMI, age during pregnancy, and social disadvantage index. Missing data were estimated using multivariate normal imputation having 25 imputations. The odds of overweight or obesity were significantly higher among infants introduced to formula or solids at ≤4 months compared to those introduced at >4 months in both unadjusted (odds ratio = 2.3262, p = 0.023) and adjusted (odds ratio = 1.9543, p = 0.0475) analyses. The odds of overweight or obesity when age at formula or solids introduction was held fixed at ≤4 months, increased significantly (odds ratio = 2.0856, p = 0.0215) for children stopping breastfeeding at age ≤4 months compared to >4 months. Thus, increasing the prevalence of breast-feeding without any formula or solids to 4–6 months in southwest Sydney should be a worthwhile public health measure.