Wikipedia:Reference desk/Archives/Mathematics/2011 July 8

Welcome to the Wikipedia Mathematics Reference Desk Archives
The page you are currently viewing is an archive page. While you can leave answers for any questions shown below, please ask new questions on one of the current reference desk pages.

July 8

Using the Anderson–Darling test

Hi there. Hopefully this is a fairly straightforward question.... I have two sets of data, let's call them D and H. D has 71 data points, H has 615. I am trying to determine which data set is the most normally distributed, and Minitab Statistical software tells me that set D has an 'Anderson–Darling Normality test' A² value of 4.27, while for H, A² = 16.90. Which of these indicates the more normally distributed set? Or, do I need a different test? Thanks in advance! --sparkl_!sm ^hey! 10:17, 8 July 2011 (UTC)

Measure the skew and kurtosis of each. Apparently, I am decades out of date on that. Please see normality test.

http://www.itl.nist.gov/div898/handbook/eda/section3/eda35e.htm suggests that set D is more normal (but that is possibly misleading. For certainty, you need to calculate which set has the greater e.g. 95% chance of being normal, I suppose. I would hope the most popular measures of normality would always vary in proportion.) 99.24.223.58 (talk) 18:37, 8 July 2011 (UTC)