Resumen

VARGAS-ROJAS, Jorge Claudio; VARGAS-MARTINEZ, Alejandro  y  CORRALES-BRENES, Eduardo. Agricultural experiments with repeated measures over time: analysis strategies comparison. Agron. Mesoam [online]. 2023, vol.34, n.2, 52634. ISSN 2215-3608.  http://dx.doi.org/10.15517/am.v34i2.52634.

Introduction. Several modeling techniques have been used to analyze experiments with repeated measures over time; however, some of these are no longer relevant. Objective. To compare four analysis strategies that are used to analyze agricultural experiments with evaluations over time. Materials and methods. Data from an experiment in which the effect of different nitrogen fertilizer sources on chlorophyll content in a forage grass at different harvest ages was used. These data were analyzed using four strategies: the area under the curve index (AUC), multivariate analysis of variance (MANOVA), random effect of the experimental unit, and temporal correlation modeling. The lasts two strategies were performed under the theory of mixed linear models; in these different models were fitted, all with the same fixed effects structure, but with different random effects, residual correlation structure, or residual variance structure. Using penalized likelihood criteria (Akaike information criterion (AIC) and Bayesian information criterion (BIC)), the best fitting model was chosen, with which inferences were made about the means of the significant fixed effects, and compared with the results obtained from the AUC and MANOVA strategies. Results. The best fitting mixed linear had a compound symmetry correlation structure and heteroscedastic variances. This model allowed for the analysis of the treatment × time interaction; on the other hand, both the MANOVA and the AUC allowed for the analysis of temporal trends of the treatments. Conclusion. The best fitting mixed linear model made it possible to select the best treatment based on the evaluation time. On the other hand, both the MANOVA and the AUC led to the selection of treatments that were not the best at all evaluation times.

Palabras clave : biometry; statistical models; analysis of variance; multivariate analysis; statistical analysis.

        · resumen en Español     · texto en Español     · Español ( pdf )