Environmental risks associated with child growth are complex, and intervention effectiveness has been consistently poor. To improve effectiveness, proper intervention points inside the complex system must be identified. Integrating site-specific knowledge, machine learning, and statistical modeling offers a powerful approach to addressing this problem. In this study, a novel four-step method is employed to identify the key environmental factors to low child height-for-age in Guatemala. The four steps included (1) the development of a region-specific, ranked list of contributing factors to low child height-for-age via informal interviews and literature; (2) the application of a clustering method to a large regional data set; (3) the identification of the top six ranked variables shared between Step 1 and 2; and (4) the analysis of the clustered, regional data set in a multigroup path analysis incorporating the top six ranked variables, diarrheal prevalence, and child height-for-age. Results suggested that an increase in diarrheal prevalence was not consistently associated with a decrease in child height-for-age. Having soap for handwashing was significantly correlated with lower diarrhea and higher height-for-age. The effect was larger in the poorer population. Finally, disease in maize was significantly correlated with lower diarrhea. This method provided an approach to reducing, modeling, and ranking large numbers of environmental risk factors to child growth, identifying potential regional intervention points.
- Cluster analysis,
- Regression analysis,
- Child height-for-age,
- Clustering methods,
- Diarrheal prevalence,
- Guatemala,
- Path analysis,
- Learning systems,
- Zea mays
Available at: http://works.bepress.com/gayla-olbricht/51/