Outlier Effect in Regression Inference Based on T Test Statistic Using Standard Deviation Method

Abstract

One of the essential requirements for smooth regression analysis based on ordinary least squares (OLS) method is normality of the data. When the dataset pass the normality test, it is virtually free from unusual observations, (OLS) method can then be applied effectively to obtain the required estimate of the regression parameter and make inference. It is quite obvious that the presence of outlier distort regression inference when non robust methods are used. This article highlighted on the consequences of outlier in the analysis of regression models based on the t test statistics for testing significant of regression coefficient, since present of outlier could disturb significant test which in turn may lead to misinterpretation of the final result. Standard deviation (SD) method were employed in measuring the effect of outlier on t test statistics taking into account the sample size and the number of regressor (s) at different intensity of outlier both using real examples and simulation study. It was discovered that outlier affect the estimate of regression coefficient negatively which in return render the classical regression estimators inefficient. In addition it can alter the odds of making both type I and type II error as well as influence the estimate of regression that are of essential interest.