[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

From |
"Visintainer PhD, Paul" <Paul.Visintainer@baystatehealth.org> |

To |
"'statalist@hsphsun2.harvard.edu'" <statalist@hsphsun2.harvard.edu> |

Subject |
st: kwallis2 and adjusted p-values |

Date |
Thu, 3 Sep 2009 11:53:14 -0400 |

A question about how -kwallis2- computes the adjusted critical level pops up occasionally (see http://www.stata.com/statalist/archive/2009-02/msg00379.html). It seems as if the adjusted critical level is too conservative for a usual Bonferroni's adjustment, using alpha/(k(k-1)) rather than alpha/k. (I even modified kwallis2 to use alpha/k adjustment). Now, however, I believe that -kwallis2- is indeed using the correct adjustment. Recently I had some data which I gave to a colleague who uses BMDP (remember BMDP?). His output was virtually identical to -kwallis2-, using the alpha/(k(k-1)) adjustment. The old BMDP manual (1992) states that the observed "Z is compared with the tabled Z[alpha*/2], where alpha* = 2alpha/(k(k-1)). So, alpha*/2 = alpha/(k(k-1)), which is what -kwallis2- uses. My interpretation of this is that the global K-W test is one-sided (e.g., a chi-square test where all squared deviations are positive), but the post-hoc pairwise comparisons are two-sided, either positive or negative deviations. Another way of thinking about it is that if the conventional Bonferroni's adjustment were applied (alpha/k), the global test would be conducted at a critical level of 5%, but the post-hoc tests would be equivalent to computing adjusted 90% confidence intervals. -p ___________________________________ Paul F. Visintainer, PhD Baystate Medical Center 280 Chestnut Street Springfield, MA 01199 ---------------------------------------------------------------------- CONFIDENTIALITY NOTICE: This email communication and any attachments may contain confidential and privileged information for the use of the designated recipients named above. If you are not the intended recipient, you are hereby notified that you have received this communication in error and that any review, disclosure, dissemination, distribution or copying of it or its contents is prohibited. If you have received this communication in error, please reply to the sender immediately or by telephone at (413) 794-0000 and destroy all copies of this communication and any attachments. For further information regarding Baystate Health's privacy policy, please visit our Internet web site at http://www.baystatehealth.com. * * For searches and help try: * http://www.stata.com/help.cgi?search * http://www.stata.com/support/statalist/faq * http://www.ats.ucla.edu/stat/stata/

- Prev by Date:
**RE: st: regression using dummy variables** - Next by Date:
**st: RE: Repeated posts are not advisable** - Previous by thread:
**st: replace a varlist to missing values for some observations** - Next by thread:
**st: simulate using lroc** - Index(es):

© Copyright 1996–2019 StataCorp LLC | Terms of use | Privacy | Contact us | What's new | Site index |