Amplitud de intervalos de confianza obtenidos por bonferroni de acuerdo al número de variables y del tama？o de la muestra en poblaciones multinormales

frequently, scientists in the field of veterinary sciences design their experimental work using multivariate analysis, which implies the selection of variables of interest and an adequate sample size. nonetheless, for different reasons, large sample sizes are not always available. an important feature of statistical inference is the estimation of parameters by confidence intervals. by using a multivariate analysis, we resorted to the use of？ the bonferroni′s inequality method, as an alternative to study confidence intervals. this method has the advantage of warranting a simultaneous confidence coefficient of at least 100 (1-α)%. the narrower the intervals are, the more accurate they will be. in this study, 3 matrixes of variances and covariance σ were used. their structures correspond to: ρij=0.25, 0.65 and 0.85; number of variables p= 2, 3, 4 and 5; and sample sizes n =10, 15, 20, 25 y 30. in reference to the interval amplitude obtained, it is concluded that as the correlation among variables or sample size increase, the intervals are narrower. finally, it was observed that when the number of variables increases, the intervals are wider.