Minimum data for performing an Anderson Darling Test

[Ed-Bal]

Hello

I have to know if a sample of data are normal or not. I want to use the Anderson Darling test, but I would like to know what is the minimum data to enter in order to get a trust result.

I know that much data is better, but there must be a minimum data which not give chance to a doubtful result

Tbnaks in advance

Eduardo Baldomero
Toluca, México

[reynald]

In Reply to Parent Post by Ed-Bal

Hello

I have to know if a sample of data are normal or not. I want to use the Anderson Darling test, but I would like to know what is the minimum data to enter in order to get a trust result.

I know that much data is better, but there must be a minimum data which not give chance to a doubtful result

Tbnaks in advance

Eduardo Baldomero
Toluca, México

Hi,
I would say between 50 to 300.
I dont have a basis other than experience.
My logic is if sample size is too small, the test would easily give a p-value of >0.05.

If the sample size is too big however, the power if the test is very high and very likely that would give p-value of less than 0.05. This is because it is very hard to find real data to be perfectly normal.

If you want to be objective about sample size though, I think someone from here can give us power curves for AD test. That way you can specify your alpha and beta to get the required sample size.