Ajustes de modelos de platô de resposta aplicados ao estudo de exigência nutricional de frangos de corte

Abstract

This work had for objective to apply the response plateau technique to the models that have maximum, to dispose the methodology and the models to be used for other researchers ofthe animal or vegetal área. The used data had been adapted from Teixeira (1994). Tne experiment was constituted by a completely randomized design with factorial structure 2x2x9, being 2 experiments, 2 sex and 9 leveis ofZn, with 8 replications each. Tne leveis of Zn were: 0, 15, 30,45, 60, 75, 90,105 and 120 ppm ofZn in the ration. Tne studied variable was levei ofZn in the tibia in ppm and levei ofZn in the médium finger in ppm. Were determined the routines ofPROC NLIN (SAS®) in accord to the methodology used for SAS®/STAT manual, with the averages ofthedoses orthe unfolding of the interaction Dose x Sex of the analysis variance initial. Two nonlinear models called MNLEl and MNLE2 and one quadratic model (MPQ) were adjusted. R2, the confidence intervals forthe parameters of the model and for the maximum dose and maximum response of Zn, and the adherence test were gotten. The R2 were raised, above of 99,5%, and well next, for ali the models and variable. The models fítted well to the data according to adherence test, a time that ali the regression deviation was not significant for test F (P>0,05). rt was obtain the variance and covariance matrix in order to estimate the confidence interval for the parameters with backup ofthe initial information, beingthe this mostaccuratethan the gotten intervalofthe adjust of PROC NLIN of the SAS®. For the maximum levei and maximum response estimated two confidence intervals, one by the algebraic method through the development for Taylor*s series and another one by means of the method jackknife, being the this last, more accurate. The MNLEl shown to be the best model among the exponentials models , also shown to be the more easyto be done. The nonlinear models are equivalents.