Modeling Experimental Designs Including Longitudinal Data and a Functional Covariate.

The study of longitudinal measures of chlorophyll concentrations is key to reducing the risk of yield-limiting deficiencies or costly over fertilizing. Factors as irrigation and fertilization can influence the chlorophyll content. In this research we analyzed data from a experimental design of chlorophyll concentrations in chili pepper plants under the effect of two factors (fertilizer and irrigation, both with four levels) recorded weekly (for seven weeks). The spectral signature curves obtained for each plant was included in the model as a functional covariate. We propose an alternative for the analysis of data from experimental designs involving longitudinal data (LD) and a functional covariate. Two smoothing approaches using basis functions and functional principal component reduce the problem to the application of a Linear Mixed Model (LMM) to LD in the presence of multiple scalar covariates. In both approaches, the results indicate that the inclusion of the functional covariate (spectral signature) contributes to explain the relationship between the chlorophyll concentration and the factors analyzed. [ABSTRACT FROM AUTHOR]