RESUMO
We use a finite population mixed model that accommodates response error in the survey variable of interest and auxiliary information to obtain optimal estimators of population parameters from data collected via simple random sampling. We illustrate the method with the estimation of a regression coefficient and conduct a simulation study to compare the performance of the empirical version of the proposed estimator (obtained by replacing variance components with estimates) with that of the least squares estimator usually employed in such settings. The results suggest that when the auxiliary variable distribution is skewed, the proposed estimator has a smaller mean squared error.
RESUMO
Predictors of attained size at 30 mo and growth rate between 18 and 30 mo (eg, diet, maternal size, morbidity, age at weaning, and selected environmental factors) were investigated by using longitudinal data from 67 Mexican children aged 18-30 mo. These children were small because of growth stunting in early infancy. Between 18 and 30 mo they grew on average at the 50th percentile of National Center for Health Statistics references values for weight, and the 25th for length. Size at 30 mo and growth rates were unrelated to energy or protein intakes during the previous year, but positively related to consumption of animal-origin foods. Maternal weight predicted size and growth only for children with larger intakes of animal products. Individual rates of weight and length growth were uncorrelated on average, correlated negatively when animal-product intake was low, and positively only when more animal products were consumed. Integrated weight and length growth, and the influence of maternal size on growth, may be disrupted by poor dietary quality. Different children would be identified as malnourished depending on whether weight or length growth was measured.