Date:         Mon, 22 Jul 1996 13:56:35 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: Probability of Cohens Kappa

In article <1.5.4.32.19960722105303.007551fc@uvapsy.psy.uva.nl>,
Harry Garst  <garst@UVAPSY.PSY.UVA.NL> wrote:
>Dear all,
>
>Oh oh oh why doesn't SPSS give me the Approximate Significance when I need it
>so badly.
>
>example from SPSS-output:
>
>                                                      Approximate
>Statistic        Value        ASE1      Val/ASE0      Significance
>
>Kappa           .76471       .08930     6.76600          ?
>
>
>Have I missed something?
 
The ratio Val/ASE0 is distributed as a standard normal in large samples,
so you can refer that to a Z-distribution if appropriate. Given this
number, even if the sample isn't very large, the ratio is large enough
to conclude that the agreement is better than chance.
 
There are a number of statistics in CROSSTABS where the significance is
not printed in some releases. This was done because approximations were
only valid in large samples and we wanted to make users make the decision
as to whether the sample was large enough. Beginning with release 6.1.2,
these are listed under the heading Asymptotic, and if you have the SPSS
Exact Tests option, you can get exact or Monte Carlo values in addition.
 
--
-----------------------------------------------------------------------------
David Nichols             Senior Support Statistician              SPSS, Inc.
Phone: (312) 329-3684     Internet:  nichols@spss.com     Fax: (312) 329-3668
-----------------------------------------------------------------------------