Resumen

GONZALEZ VILLALOBOS, Decio Martín et al. Analysis of Three Statistical Procedures to Evaluate Growth in Crossbred Heifer Under Different Nutritional Levels.. Rev. Cient. (Maracaibo) [online]. 2007, vol.17, n.2, pp.136-142. ISSN 0798-2259.

There are investigations where variables are measured in periods of time on the same animal. This type of information should be analyzed statistically trough three ways: univariate analyses with the RANDOM statement of the GLM procedure; univariate or multivariate analysis with the method of lineal transformations with the REPEATED statement of the GLM; and with mixed models of covariance with the MIXED procedure. With the objective of evaluating these three statistical methods and to know the most precise, biweekly live weight coming from a rehearsal carried out located in the Tachira State, Venezuela (topical damp woods) was analyzed during 7 weeks, where 30 crossbred heifers with an average weight and age of 176.9 ± 24.6 Kg and 17.22 ± 2.23 months respectively, were randomly distributed between three groups: (1) control, (2) balanced commercial feed, and (3) Flour of Gliricidia sepium with flour of corn and molasses. It was modeled covariance structures, comparing the GLM procedure with its RANDOM and REPEATED statements vs. the MIXED procedure in its CS, UN and AR1 options, of the statistical package SAS. As dependent variable it was studied the weight of the heifers during the assay period and as independent variables the supplementation group, the period and the linear interaction among both. When carrying out the analysis of variance using the most suitable structure of errors, it can be conclude that there was a significant interaction between treatment and period (P<0.01), and that is to say that the curves of growth spread unless parallel. Results indicate that the best fitting analysis is the Proc MIXED with the AR1 option, since it allows to adjust the covariance womb.

Palabras clave : Growth; repeated measures; statistical comparison; heifers.