Talk:Hartley's test
 | This article is rated Stub-class on Wikipedia's content assessment scale. It is of interest to the following WikiProjects: | ||||||||||
| |||||||||||
Reference to Cochran's C test modified
[edit]'The previous text refrered to a "test due to Cochran". I have made the reference more specific and I added a cross-reference to the article Cochran's C test. --Rtlam (talk) 09:37, 22 September 2011 (UTC)
Warning About Errors in the Cited Table of Critical Values for Hartley's Fmax Statistic
[edit]The cited table of Fmax critical values is not completely reliable.
The .05 critical value for 2 treatments and 7 degrees of freedom (df) should read 4.99 rather than 0.99; the .01 critical values for 3 df for 7 to 12 treatments are off by at least a factor of 10. For these latter cases the Biometrika Tables for Statisticians noted that their least significant digits might be uncertain by several units and (for example) wrote the entry for 7 treatments as 21(6) rather than as 216 to alert the user to that possible imprecision; the cited table ignores the final parenthesized digit all together, writing 21, which is grossly wrong. The .05 critical value for 2 df and 12 treatments is given as 704 in the Biometrika Tables and in Lloyd S. Nelson (1987) "Upper 10%, 5% and 1% Pointsof the Maximum F-Ratio". Journal of Quality Technology 19(3), pp. 165-167; the cited table gives 714.
Fortunately, few of these critical values are likely to be used in practice by statisticians since they have only 2 or 3 degrees of freedom or (in the case of 2 treatments) a standard F-table would be used instead.
Unfortunately, I haven't computed these tables myself, having relied on copyrighted sources, and therefore can provide no uncopyrighted sources for Wikipedia. --Ray921 (talk) 21:58, 30 August 2013 (UTC)