Date:         Mon, 10 Mar 1997 15:01:43 GMT
Reply-To:     David Nichols <nichols@SPSS.COM>
Sender:       "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From:         David Nichols <nichols@SPSS.COM>
Organization: SPSS, Inc.
Subject:      Re: Noncent. Parameter

In article <AF45998F-10452@206.86.233.115>,
Jamin Brazil  <jamin@BEST.COM> wrote:
>I have run a table titled Tests of Between-Subjects Effects of which
>there is a banner point called Noncent.Parameter.  I cannot figure out
>what it means...can anyone please tell me or point to to the right
>materials?
>
>Thanks very much,
>Jamin
>
>---------------------------------------------------------
>Jamin Brazil  =80  jamin@best.com  =80  http://www.macroinc.com
>---------------------------------------------------------
 
The noncentrality parameter is used in calculating the observed power
value given next to it. Numerically, for univariate F-tests, the
noncentrality parameter is calculated as SSH/MSE, where SSH is the
hypothesis or effect sum of squares and MSE is the error mean square.
The observed power takes the given data as representative of the
population and estimates the proportion of times a significant
result would be obtained in a random sample of this size. The
noncentrality parameter indexes the departure of the data from the
null hypothesis of no effect. If the size of effect seen in the data
is indicative of the population effect, the distribution of the
F-statistic will not be given by a central F, but rather by a
noncentral F with a certain noncentrality parameter. This parameter
is what is printed.
 
--
-----------------------------------------------------------------------------
David Nichols             Senior Support Statistician              SPSS, Inc.
Phone: (312) 329-3684     Internet:  nichols@spss.com     Fax: (312) 329-3668
-----------------------------------------------------------------------------