CMStatistics 2022: Start Registration
View Submission - CMStatistics
Title: Estimation of dynamic diarrhea effects on childhood growth with latent subgroup Authors:  Tonghao Zhang - University of Virginia (United States)
Jianhui Zhou - University of Virginia (United States) [presenting]
Jennie Ma - University of Virginia (United States)
William Petri - University of Virginia (United States)
Abstract: Existing studies treat diarrhea effect on childhood growth as constant over time, and also assume the effect is the same across the population. However, in practice, diarrhea episodes may have different clinical severity depending on the types of viruses infected and on the diarrhea onset time. Moreover, the individual reaction to a diarrhea episode may vary across children due to their different health characteristics. As a result, the diarrhea effect on children growth is heterogeneous across the cohort and is dynamic across time. We propose a semi-parametric model with latent subgroup to model the heterogeneity and dynamics of the diarrhea effect. To accommodate the heterogeneity of the effect, natural growth and diarrhea effect are modeled as random effect curves and the mixture distributions are adopted. The latent subgroup in the mixture distribution model helps to explain individual characteristics such as household socioeconomic status, hygiene level, and other driving factors behind diarrhea. Simulation studies show that our model achieves simultaneous identification of subgroups in growth pattern and diarrhea vulnerability and estimation of the dynamic diarrhea effect curves. The estimator variance is also validated by empirical coverage probability. The proposed model is applied to a dataset collected from children in Bangladesh, to illustrate the application of the proposed methods.