Abstract

RAMIRO VASQUEZ, Edison  and  CABALLERO NUNEZ, Alberto. When the assumption of homocedasticidad in variables with binomial distribution fails. cultrop [online]. 2011, vol.32, n.3, pp. 191-199. ISSN 0258-5936.

The process of Monte Carlo simulation was used to generate populations of random variables with Binomial distribution with homogeneous or heterogeneous variance; for 5, 10 and 30 observations by experimental unit (n) and likelihood of success of the event of 0.10, 0.20,..., 0, 90 (p). It consisted of experiments in randomized block design with 3, 5 and 9 treatments (t); 4 and 8 replicas (r); for each combination t-r-n, generated 100 experiments. By way of a reference point of discussion to have, included the variant of other 100 experiments of variables with Normal distribution, with similar mean and variance of the experiments of data with Binomial distribution. It was found that the behavior of the indicators: percentage of experiments in which there is a rejection of the hypothesis H0; the power in the ANOVA; minimum difference detected in the experiment, as well as the number of differences between treatments is similar within each alternative analysis through three variants; showing a marked influence the number of observations per experimental unit and the number of replicas in these indicators.

Keywords : homogeneity of variance; homocedasticidad; variables binomials; assumptions of ANOVA.

        · abstract in Spanish     · text in Spanish     · Spanish ( pdf )