PDA

View Full Version : Anderson-Darling test?



hcrisp
09-15-2009, 02:48 PM
The NORMALITY function in the IMSL Statistics Reference only includes the Shapiro-Wilk W test or the Lilliefors test for normality. Is there a function for performing the Anderson-Darling normality test?

totallyunimodular
09-15-2009, 04:24 PM
There is no test for Anderson-Darling. That being said, Shapiro-Wilk is generally seen as preferable to Anderson-Darling (as powerful, more robust). Is there a particular reason you want to use AD?

hcrisp
09-16-2009, 07:17 AM
Only that it is available in Minitab and I am trying to move a user's current process from Minitab into PV-WAVE.

totallyunimodular
09-17-2009, 09:55 AM
Well, luckily the AD statistic is straightforward to compute. For example, see the Wikipedia write up (http://en.wikipedia.org/wiki/Anderson%E2%80%93Darling_test).

You can use the PV-WAVE standard library routine SORT to sort the data, and then the Statistics Toolkit routine NORMALCDF as the other key component.