Talk:Q test

From Wikipedia, the free encyclopedia

This article is within the scope of WikiProject Statistics, which collaborates to improve Wikipedia's coverage of statistics. If you would like to participate, please visit the project page.

Does anyone have a particular reason why the Q can only be used once per data set? I understand it's not necessary when one data point has experimental error, but couldn't it be used on two different points if there still enough degrees of freedom left after the first point is rejected?

I'm a chemist, not a statistician, so any clarification would be great, especially since this page can then be updated.

Sorry, I am not a Wiki expert, so excuse me if I have done this incorrectly. The first line of this article seems to be a copy and paste from http://science.widener.edu/svb/stats/qtest.html

[edit] Article Quality

I'm still a new editor, but it seems like this page needs a bit of work. As another person stated, it looks like the first line is a copy and paste, and it fails to really describe what a Q-test is, and where and why it would be used. What is gap, in the equation? Bigdavesmith 22:06, 26 June 2007 (UTC)

The "gap" as it is defined here is the difference between the datapoint in question and the datapoint whose value is closest to the suspect outlier.

There should be more examples of Q-test to understand properly. By Pavan Talwani

[edit] Why use only once

The test for throwing out the outlier deals with the gap between the other values and the outlier. Its intent is only to be used once, per the author's instructions on how to apply it (see below). The tables only deal with 2 to 10 values, so a data set with two outliers in just ten values sounds like more work needs to be done to get consistent data. This test should actually be called Dixon's Q test, after the author of the original paper in Analytical Chemistry. Simplified Statistics for Small Numbers of Observations R. B. Dean and W. J. Dixon Anal. Chem., 1951, 23, (4), pp 636–638 Hoffrick 00:46, 31 August 2007 (UTC)