Date: Fri, 10 Sep 2004 10:01:09 +0200
Reply-To: Henrik Lolle <lolle@socsci.auc.dk>
Sender: "SPSSX(r) Discussion" <SPSSX-L@LISTSERV.UGA.EDU>
From: Henrik Lolle <lolle@socsci.auc.dk>
Subject: Re: Ordinal scale and using mean
In-Reply-To: <20040908130420.9470.qmail@web60509.mail.yahoo.com>
Content-Type: text/plain; charset="us-ascii"; format=flowed
Hi Paul,
I think that a nonparametric statistic like Mann Whitney would do here. And
I don't think that changing the values will make any difference in the
result, as long as it is the same changes to both samples.
But it confuses me that you write 'two variables' which indicate that your
samples are related (two questions to the same set of respondents and not
one question to two differenct samples of respondents). In that case the
independent sample tests does not work very well. In SPSS you can then move
to Wilcoxon or Sign test for paired samples.
Henrik
At 15:04 08-09-04, Paul Hover wrote:
>Hi all,
>
>
>
>I have two ordinal variables X and Y with 5 values each: from strongly
>disagree -> strongly agree. I would like to compare these two variables by
>calculating a mean for each variable. This implicates that every possible
>answer must be assigned a number, e.g. 1 to 5 or -2 to +2. Fact is that
>there are, of course, different outcomes!
>
>
>
>You can compare the % 'strongly agree' for X and Y (and so on), but then
>it's difficult to make a concise conclusion. Another option is to do Mann
>Whitney or Kolm. Smirn. test (but still then: with which values/scale?).
>
>
>
>Any suggestions what to do?
>
>
>
>Thanks!
>
>
>
>Paul
>
>
>
>
>
>
>
>---------------------------------
>Do you Yahoo!?
>Yahoo! Mail - You care about security. So do we.
***********************************************************
Henrik Lolle
Associate Professor
Department of Economics, Politics and Public Administration
Aalborg University
Fibigerstraede 1
DK 9220 Aalborg East
http://www.socsci.auc.dk/institut2/dansk/empl/lolle.htm
***********************************************************
| 
