I chose colors of the visible spectrum precisely because the are
ordered and do correspond to frequencies; they work just as well as
1-5 here.
LOL, by your definition any data set can be considered ordinal since you
could "order" any choice in some bizare way. Any set of chocies can be
arranged Alphabetically for instance. Clearly that makes no sense.
The difference between choosing the favourite color, and choosing the
degree of "on-topicness" is pretty clear , if you think about it.
In one scale, a voter choosing 3, would strictly imply he prefers 3 to 2
and 2 to 1. In your color example, a voter who prefers Yellow, tells you
that he prefers yellow, period.
The numbers 1-5 were chosen as a nominal "scale", as
Susan has made clear.
That might or might not be the intent, but if you look at the sample quite
a few people chose 0. That makes no sense if it was a nominal scale.
It would make ,some (admitedly not much) sense if people saw it as a
ordinal scale.
Any attempts to operationalize "usually" and "sometimes" would IMO
be a waste of time, if only because there would be no concensus on
how to do so.
I never said it would be easy. An ordinal scale is the best you can do. A
Likert scale or something similar.
Maybe so. I'd argue that the two are qualitatively different
choices but they are not differentiable in a quantifiable way,
The choices can be qualitatively different and still be ordered. For a data
set to be catergorically (either ordinal or nominal), all that is required
is for the choices to be discrete, in some cases (ordinal), the choices can
be ordered.
I don't believe being able to quantify your choices exactly is a necessary
condition to ordering them. Look at the Likert scale again.
In any case, if the choices were meant to be nominal
(red,yellow,green,blue) the order of the choices and the wording given was
misleading , because they were arranged in what one might expect to be in
order of severity from most on topic to least on-topic.
1. off-topic - discuss only when a warning is needed
2. off-topic - brief mention sometimes okay (for comparison etc.)
3. ???-topic - sometimes okay to discuss
4. on-topic - usually okay to discuss
5. on-topic - always okay to discuss
Notice the frequency of "okay to discuss" increases. This looks like a
ordinal scale to me, rather then something closer to your color example.
then
I'd use that to back up my assertion that these things were not put
on a meaningful 'scale' in the first place, that the labels were
nominal, not ordinal. If I had time. Heh.
Running away , I see
That's what I meant to convey when I typed "does not have any
bearing."
No, I'm wondering why you are bashing on our friend the "mean" with your
story even though sensitivity to extreme values isn't really important
here.
Aaron (my email is not munged!)
