THE EFFECTS OF THE VIOLATION OF ERROR'S INDEPENDENCE HYPOTHESIS IN TWO-WAY MODELS.

Luiz Carlos BAIDA[1]

Antonio Francisco IEMMA[2]

 

§    ABSTRACT: This is a study of the implications induced by violation of the independence assumptions of types I and II errors probabilities, inherent to classical hypothesis in the linear models with fixed and balanced two factors, analyzed under the standard structure of Gauss-Markov's linear models. The theory is developed in an appropriate form, studing three equicorrelation cases of the errors of the model. Tables of the errors probabilities are given to show the bias that appears when the standard model is used inadequately. To produce these tables we have considered various values of the level of the factors, the size of the samples, the involved correlations and finally, the pre-fixed nominal level of the significance (( = 0,05).

§    KEYWORDS: Types I and II errors probabilities; analysis of variance; dependent errors; equicorrelation.



[1] Departamento de Ciências da Computação e Estatística - Instituto de Biociências, Letras e Ciências Exatas - UNESP - 15054-000 São José do Rio Preto - SP

[2] Departamento de Matemática - Escola Superior de Agricultura " Luiz Vaz de Queiróz" - ESALQ - USP - 13418-260 - Piracicaba – SP.