Testing Normality of Data using SAS

Many statistical tests require data to be approximately normally distributed. Usually, the first step of data analysis is to test the normality. Also, we often test the normality of residuals after fitting a linear model to the data in order to ensure the normality assumption of the model is satisfied. SAS has offered four statistical tests that provide an easy way to test the normality. However, we should be cautious when we use these tests due to their limitations. In some cases, we may draw incorrect conclusions by only looking at the test statistics and p-values. Graphical methods are powerful in displaying distribution characteristics of the data and can serve as a useful tool in checking the normality. Combining graphic methods and statistical tests will improve our judgments on the normality of the data. In this paper, I will present these methods SAS uses by applying them to the real data from a clinical trial.