Re: Ordinal Outliers?

[David Stodolsky <david.stodolsky@socialinformatics.org>]

I don't see how to compute a non-parametric similarity (or distance)  
measure. First problem is that it seems that SOMERS' D is the  
appropriate procedure (I want to avoid Tau B, since the table isn't  
square) and then there doesn't seem to be any way to save the results,  
so I can feed them into a MDS or factor analysis.
dss


On 11 Jul 2009, at 9:23 PM, James Gundlach wrote:

> In general ordinal variables do not have outliers and the reason is  
> that
> outliers are a statistical problem in continuous or interval data.  If
> you double click the column heading a variable information box opens  
> and
> if you change the Modeling Type from Continuous to Ordinal JMP will
> automatically select the kinds of statistics appropriate for your
> variables.  You will simply have to pick from those produced to see
> which is appropriate for your data.
>  jim gundlach
>
> > > > David Stodolsky <david.stodolsky@socialinformatics.org> 07/11/09
> 1:28 PM >>>
> I have some ordinal data (5 levels) that I need to eliminate outliers
> from before doing MDS. The older version of JMP I have, requires
> continuous data for Man. D or Jackknife. Is there some other way to do
> this or do I need some other program? The data contains two subsets of
> questions (attitudes and what could change attitudes) and the
> population may be heterogeneous.
>
>
> dss
>
> David Stodolsky, PhD                   Institute for Social  
> Informatics
> Tornskadestien 2, st. th., DK-2400 Copenhagen NV, Denmark
> dss@secureid.org                                      Skype:
> davidstodolsky
David Stodolsky
dss@secureid.org                                      Skype:  
davidstodolsky