Pendekatan Regresi Berganda pada Analisis Varians Klasifikasi dua Arah

Abstract

The procedurs for multiple regression to obtained a solution b = (XT X)-1 (XT Y), of the normal equation XT Xb = (XT Y), it is necessary that XT X be a nonsingular mattrix. All this means in practise is that the normal equations must involve as many independent equations as there are parameters to be estimated. If data are obtaned from a designed experiment, however, some case is needed to check that all the normal equations are independent, or if they are not, take steps to obtained stimated just the same. Multiple Regression approach can be used for solving Analysis of Variance problem, whether of one way or two ways ANOVA. In ANOVA, model is an important factor. This study focus on two ways classifications ANOVA with regression approach for solving the fixed effect model. it can be done if the model is identified truly and if the preventive procedure have been done so an independent normal equation has been.A characteristic of ANOVA is that the analysis model is overparameterized, so have to made constraint for parameters. In the Multiple Regression model approach for ANOVA problem, the independent variable X at categorical from 0 and 1 or have to make dummy variables at row and colomn factors.