A study on testing regression parameters under heteroscedasticity

In this study, on a general linear model with unknown and possibly unequal normal error variances, One-sample sampling procedure offered by Wen, Chen and Chen (2006) is described to handle the hypothesis testing on all, partial or a subset of linear functions of regression parameters. The procedure can be applied to analysis of variance under various designs of experiments with unequal variances. The offered procedure and the estimators of regression parameters are desribed in second chapter. In third chapter hypothesis testing on all slope parameters or a subset of regression parameters, its application to analysis of variance and an extension to general linear hypothesis are studied. In fourth chapter, offered test statistic F ~ and usual F test results are compared with data derived from simulation program, an application to general linear hypothesis is given and finally the power of F ~ and the power of usual F are compared with a simulation study. Key Words : Regression analysis, heteroscedasticity, one sample procedure, hypothesis testing, ANOVA