Anderson Darling normality test

To check if the data-set is distributed according the normal distribution. Various statistical test assume that the data is normally distributed therefor it is important to check this. If the data is not normally distributed looking at the Kurtosis and Skewness and Histogram of the data gives an indication of the shape of the data.

Tests that assume a normal distribution

• one sample t-test
• 2 sample t-test
• Paired t-test
• Variation F-test
• One way Anova

Interpretation

• The smaller the p value the more unlikely the data-set is normally distributed.
• In Develve the minimum p value for normality is 0.10.
• The minimum advised sample size for this test is 20 samples.
• Alway check the Histogram of the data-set.
• What to do with not normally distributed Data

Colors of the cells

• Red
  Data-set is not normally distributed
• Orange
  Sample size to small

Normal distributed data set

Not normal distributed data set

Formula

Correction factor for small sample sizes.

With the A² value can the program interpolate in a table the p value.

Legend

n = n

Example

Data-set C is not normally distributed (normality p <0.10) therefore the cells of the statistical test that assume normality are colored red.

Data file

External links

• http://en.wikipedia.org/wiki/Anderson%E2%80%93Darling_test
• http://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm
  

To check the normality of a data-set.