How to test normal distribution

How to post

You need to resisger

Get user name and password

Hit new thread

type your tile

type your question or comment

hit post

easy, daniel:)
There are several methods that are generally considered to give accurate results. Often as a first step, the data are plotted as a histogram and the fitted normal distribution is plotted over the data. You could also construct a q-q plot that plots the observed quantiles with the quantiles of the fitted normal distribution. In each of these cases, the 'normality' of the data is assessed visually. For a more formal 'test' of normality, there are the Anderson-Darling and Shapiro-Wilk tests. Most statistical software packages are able to compute these tests.

The Shapiro-Wilk test is implemented in R via the following command:


Some statisticians argue that visual inspection is sufficient to make the 'normality' assumption. However, tests such as the Shapiro-Wilk are frequently used.

normal distribution test

Hi Matt
I don't think that I will plot the graph and visual check. I will use software
for Shapiro-Wilk test. Do you know which stat software which is easy to use. SAS ? Systat ? SPSS?
Currently I have SAS JMP in my desktop. I will explore more to check this software.
Thanks million for your valuabel advice
Currently I have SAS JMP in my desktop.
Analyze -> Distribution
Select your response, add it to the Y, Columns box, OK
Click red triangle next to the response -> Fit Distribution -> Normal
Click red triangle next to Fitted Normal -> Goodness of Fit

This is how the Shapiro-Wilk W test is done, for small data sets.