SLRMss: um pacote em R para a correção de testes em modelos de regressão linear simétricos

Abstract

There are regression models to model functions which have random errors with symmetric distributions. In order to test this model parameters, two hypothesis are made, H0 and H1, in the first one, the case which a vector of regression coefficients are equal to a fixed value is considered, in the second one, the case of at least one coefficient is different than the fixed values is considered. The test statistics have, under the null hypothesis, approximate distribution chi-squared. For large samples, this approximation has a small approximation error, but, for small samples, this approximation may show distorted results, being needed corrections. This work shows the creation process and usage of the SLRMss (Symmetric Linear Regression Models for small samples) package, a package in the language \faRProject\ which computes estimates and hypothesis tests for symmetric linear regression model with small samples. Four distribution were used to model the errors distribution, normal, t-Student, power exponential and logistic II and four statistics were used to test the significance of parameters, Wald, likelihood ratio, score and gradient, for the last three of them, their correct version were also computed. The steps to create the package and the application to real data available in the package are also present in this work.