Dear Ulrich,
Thanks for your reply. I have obtained a book of Agresti from library --
Statistical Methods for the Social Sciences, 2nd edition. I fortunately
locate Section 7.4 titled 'comments concering nonparametric statistics'
(p. 186), which explictly addresses the problem of treating ordinal
variables as interval variables. It mentioned that the scoring system
assigning scores 1, 2, 3, 4, and 5 to the five categories, say, very
liberal, slightly liberal, moderate slightly conservative, very
conservative, is quite dangerous...
But from your text, you incline to deem 5-point scale as an interval
scale. I wonder if there is any research which explicitly address this
question? I mean those researches which show that using interval scale
analysis is all right. If so, how about 3 or 4 points scale?
By the way, I don't find text related to the alternative scorings for
ties in Agresti book. Could you give me a more specific reference, say
book name.
Wen-Feng
-----------
In article <[EMAIL PROTECTED]>, [EMAIL PROTECTED]
says...
>
> If a 5-point scale was created or adapted for a study with any
> attention to detail, then it will be pretty close to "equal interval"
> by intention, and probably by result.
>
> If a scale has only 5 points, then there will be a whole lot of ties
> when you apply the rank-order transformation. If you read your
> statistical texts closely, you will see that they raise some doubts
> about using ranks when there are ties. Agresti has a good example of
> alternative, competing scorings for a few derived categories, in each
> of his books on categorical analyses.
>
> I think you have been exposed to the prejudices of a few
> Experimentalists in psychology and education, who were
> overly-impressed--for a little while, about 30 or 40 years ago--with
> the miracles of "nonparametric analyses which don't need any
> assumptions!" A 5-point scale happens to be "ordinal" which is a
> word that impressed a few people; but it is not well suited to
> rank-transforming (the ad-hoc solution for ties is not great).
>
> There were few (no?) credible sources that ever recommended ranking
> the 5-point scores, as far as I know--and I have asked about this
> before. So, statisticians have not considered the question especially
> noteworthy. (And, I am curious, Do recent Experimental Design texts
> say anything?)
>
>
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
