On 31 Jan 2000 09:11:58 -0800, [EMAIL PROTECTED] (Jan de Leeuw)
wrote:
> I am glad this one is back after a short absence. Likert scales (or
> any other data) "are" not ordinal or interval. Actually, they have no
> idea what they are. It's up to you. You choose the statistical
> technique and you must be able to defend your choice to your committee,
> to your client, to the reviewer, to the judge, to your students.
> The defense "I am doing this because books A,B,... tell me to do so"
> simply will not fly in some of these scenarios (and should not fly in any
> of them).
...
There is the question, "What do *you* mean by Likert?"
I have seen people use the term -- writing to the stats-groups -- when
they meant "additive", from a set of dichotomous items. That is far
inferior to having a set of "agreement" items, each scored on a
symmetric scale of 4 to 7 points, with each mean near the middle, and
similar standard deviations.
If you meet the stringent conditions of the original Likert scales,
you're dealing with interval, not just ordinal, insofar as you can
judge from good intentions, and the usual conventions of psychology.
I thought that the question (Is this merely ordinal?) usually came up
about the analysis of Likert *items* rather than Likert totals.
Well, any items with just a few, ordered categories are pretty *safe*
to treat as interval -- there can't be outliers to screw up the
variance, which is the main risk of "non-normality". To be really
concrete, you can check out precisely what scores your rank-transform
gives you for a set of frequencies of categories, by creating that
many "ties" for each, and figuring the average; then translate those
by linear algebra to scores of {low=1.0, next-low=2.0, next_a= 2.2,
next_b=2.3, next_c=2.5} -- or whatever, to see how the scoring
compares to {1,2,3,4,5}. Agresti's textbooks give a detailed example.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===========================================================================
This list is open to everyone. Occasionally, people lacking respect
for other members of the list send messages that are inappropriate
or unrelated to the list's discussion topics. Please just delete the
offensive email.
For information concerning the list, please see the following web page:
http://jse.stat.ncsu.edu/
===========================================================================
