Date: Fri, 3 Jul 1998 01:09:09 -0500
Reply-To: "Nichols, David" <nichols@SPSS.COM>
Sender: "SPSSX(r) Discussion" <SPSSX-L@UGA.CC.UGA.EDU>
From: "Nichols, David" <nichols@SPSS.COM>
Subject: Re: QUERY: estimating the true preferential order of objects
Ties can be handled, and SPSS will do so in Release 6.1.4, but you do
need complete data.
Principal Support Statistician and
Manager of Statistical Support
>From: Jayasinghe, Gayathri[SMTP:G.JAYASINGHE@CGNET.COM]
>Sent: Thursday, July 02, 1998 6:44 PM
>Subject: QUERY: estimating the true preferential order of objects
>I am planning the analysis of a questionnaire survey in which respondents
>give priority rankings to a selection of objects. eg,
> Respondent type1 type2 type3 type4
> 1 3 1 2
> 2 1 3 2 4
> 3 2 1
> 4 1 1 2
>Note: Ties are possible. Less than the max number of objects may be ranked.
>I need to,
>1/ estimate the preferential order of objects overall.
>2/ test whether the preferential order of objects is different between
>different categories of people, for example, between males and females,
>between different age groups..etc
>I have just one reference which describes Kendalls coefficient of
>concordance which then goes on to suggest an estimate for preferential order
>of objects as simply the ranking of the column sums. But this assumes no
>ties and in the case of incomplete rankings, that each object is ranked by
>an equal number of respondents (ie a balanced design). In the unbalanced
>case would it be wrong to divide the column sum by the total count of
>respondents giving a rating for this object??
>Any references to literature, software, robust methods and /or pointers and
>advice much appreciated. I do not have access to a good statistical library
>unfortunately, so any direct help would be immediately most useful.
> Statistical software available: SPSS 6.1.4 with Professional and Trends.
> Please reply to
>International Irrigation Management Institute
>Thanks in advance.