Longitudinal principal components analysis of patterns and predictors of growth in Guatemalan children.

Longitudinal principal components (LPC) analysis was used to assess growth patterns in children from rural Guatemala in order to determine if this methodology could provide additional information regarding correlates of growth compared to more traditionally used methods based on attained size and increments. LPC analysis reduces measures at many points in time into a few parameters. However, LPC analysis requires complete data, and many cases may be lost due to missing values. Thus the potentially greater sensitivity of LPC analysis should be weighed against the reduced power resulting from smaller sample sizes. Component indices representing centile level and centile shift, attained size, and 3 to 36 month increments of growth in length and weight were used as the dependent variables in multiple regression models in order to examine the effects of environmental variables, such as home dietary intake, supplementation, and prevalence of diarrhea on growth. Regardless of which growth index, i.e., attained size, incremental change, or principal component, was used, regression results were similar; higher nutritional intakes were generally associated with greater and more rapid growth from birth to age 3 years. The possible advantages of LPC analysis over more traditional methods were not great; therefore, LPC analysis is not recommended as the method of choice in this population.