home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!utcsri!skule.ecf!torn!nott!dgbt!netfs!ub!zaphod.mps.ohio-state.edu!darwin.sura.net!jvnc.net!yale.edu!news.yale.edu!mars.caps.maine.edu!maine.maine.edu!halteman
- From: HALTEMAN@MAINE.MAINE.EDU
- Newsgroups: bit.listserv.stat-l
- Subject: Re: Normal probability plots
- Message-ID: <92323.163557HALTEMAN@MAINE.MAINE.EDU>
- Date: 18 Nov 92 21:35:56 GMT
- References: <STAT-L%92111709503369@VM1.MCGILL.CA>
- Organization: University of Maine System
- Lines: 16
-
- With regard to using Pearson's Chi-Squared for testing normality,
- there is an article in the _Anals of Math Stat_ (V25) by Chernoff
- and Lehmann that shows if cell boundries defined by using maximum
- likelihood estimators based on ungrouped data (i.e our usual
- xbar and s2) , the resulting statistic is not Chi-Squared, but
- is stochastically bounded between chi-squares with (k-p-1) and
- (k-1) degrees of freedom. (k is number of cells and p is number of
- parameters estimated).
-
- Although I don't know which test is "best", it seems to me the
- usual Pearson is not the one.
-
-
- Bill Halteman
- Dept. of Mathematics
- U. of Maine, Orono
-