Statistical modelling of sugar cane stems emergency

Abstract

Growth curve analysis was used to explain the emergence of stems of two varieties (CP 65-357 y LCP 85-384) of healthy sugar cane seed obtained by two methodologies: micropropagation and hydrotermotherapy. The aim of this study was to model the effects of varieties, methodologies used and the interaction between them on the evolution of the number of stems per meter. Three replicates per variety and methodology were carry out and, for each replicate, the plot was considered as the test unit. From the descriptive analysis of the data, a logistic regression fit was selected. Initially, fixed models were applied, but due to the non-constant variance of the data nonlinear modeling with random effects for longitudinal data was applied where the plot was considered to be the unit or subject. Parameters corresponding to the horizontal asymptote and the average value of the horizontal asymptote were considered as random effects in order to reduce variability. The results showed significance in the logistic regression parameters. The horizontal asymptote and the number of days necessary to reach 50 % of maximum emergency were significant for variety, while the methodology and interaction were not significant. The assumptions of normality and constant variance on the residuals were analyzed and turned out to be adequate. As a conclusion, it can be proposed that the logistic regression model adequately explain the behavior of emergency data stems per meter.