Portal de Periódicos da CAPES

Artigo Produção Nacional

BIBTEX RIS

ANÁLISE DAS PRESSUPOSIÇÕES E ADEQUAÇÃO DOS RESÍDUOS EM MODELO DE REGRESSÃO LINEAR PARA VALORES INDIVIDUAIS, PONDERADOS E NÃO PONDERADOS, UTILIZANDO PROCEDIMENTOS DO SAS®

2011; UNIVERSIDADE FEDERAL DE SANTA MARIA; Volume: 33; Issue: 2 Linguagem: Inglês

10.5902/2179460x9359

2179-460X

Janete Pereira Amador, Sidinei José Lopes, João Eduardo da Silva Pereira, Adriano Mendonça Souza, Marcos Toebe,

Geography and Environmental Studies

It is appropriate to use regression analysis establish relations that allowto predict tone or more variables in terms of others. When there arerepeated measurements for independent variable X for differentmeasurements for dependent variable Y, the regression model may beadjusted in three different ways: using individual values of X and Y(considering all data); with means of Y for levels of X (treatments) and,using weighted means of Y by the number of repetitions of each level ofX (treatment). The objective of this study is to adjust a linear regressionmodel by individual values with weighted and not weighted means ofthe treatments in order to test the presuppositions for the adequacy ofthe model and to analyze the variance decomposing the sum of squaresof error in its components, thus evaluating the Lack of Fit. Theadjustments of the models and its presuppositions were done in SAS.Thus, it was observed that the adjusted models for individual data andweighted means present the same coefficients. The test for Lack of Fit isonly possible with individual data. The choice of best strategy to analyzethe data should be decided by the researcher but it is suggested that,when all data of the research are accessible, the best strategy would beto estimate the model using individualized data since it presents moreprecise information regarding the variability of the data set which doesnot happen when working with means of variables.