Alternativas para análise estatística de experimentos envolvendo densidades de aves, programas e fontes de luz, em matrizes de frangos de corte

Abstract

In agricultural experimentation and mainly poultry farming, it is common that in the estabhshment of the experiment, some ractors present restrictions in randomization. This fact occurs when one has to evaluate ractors such as Ught programs and sources. The present work is designed to present some alternatives to the statistícal analysis for egg production, comíng from an experiment, in which the ractors program and source oflight and population density ofbirds were evaluated, underthe sítuation in which the lightprograms andsources presented restriction in the randomization. The schemes ofstatistícal analyses approached were: ractorial, splitplot, spht splitplot and strip-plot. For each scheme routines suítable forthe variance analysis and unfoldíng ofthe interaction byusing procedures ofthe SAS® software. It was found that as the greatest restrictions are used , the degrees offreedom ofthe residue decrease, becoming more diffkult to detect significant dififerences among the means ofthe treatments. In experiments with restriction in the randomization, the comparisons among the leveisof the ractors aswellasthe studyofinteraction involve the linear combination ofone or moreerrors present in the model. In the data analysis oftotal egg yield, it was verified that only the main effects ofthe ractors were significant. As the restrictions were considered, there was a reduction in the degrees offreedom ofthe residue and hence an increase ofthe estimate ofthe experimental error.