Date: Fri, 8 Mar 2002 09:14:14 +0000
Reply-To: David Hitchin <D.H.Hitchin@sussex.ac.uk>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: David Hitchin <D.H.Hitchin@sussex.ac.uk>
Subject: Re: Reporting ANOVA results
Content-Type: text/plain; charset=us-ascii
The question was specifically about what should be shown in a table, but at
some point there should have been a check that the data values (within
cells) were reasonably close to a normal distribution. The ONEWAY command
doesn't provide facilities for doing this directly, but it is easily done
with other commands, for example
GET FILE='C:\Program Files\SPSS\Employee data.sav'.
UNIANOVA salary BY educ /SAVE = RESID.
GRAPH /HISTOGRAM=res_1 .
which uses the familiar "employee data" from the SPSS examples, and the
plots show clearly that the distribution is so heavily skewed (as often
happens with salary data) that the computed p-values will be affected.
In my opinion everyone doing an anova should run a similar check. If severe
problems of skewing exist, then the data might be transformed, care taken
to deal appropriately with outliers, and the final plots should look right.
Most people wouldn't make such plots a formal part of a presentation, but
would have them to hand, just in case someone asked "Is your data close
enough to normal for the anova to be valid?"
--On 07 March 2002 19:33 -0500 Jessica Kenty <email@example.com>
> Hi everyone,
> I have a question about presenting data. I have to give a talk this
> weekend and my analysis just focused on comparing means using a One Way
> ANOVA. I was wondering what data must be presented in a table to meet
> the basic standards of professionalism. I mean, I know I want to present
> the means & the level of significance, but what else should I include?
> What might people question me about methodologically when using ANOVA to
> compare means?
> Thanks for your help in advance!
> Jessica Kenty
> Research Assistant
> Assets & Educational Inequality Project
> Northeastern University
> Boston, MA