On 26 Jan 2001 09:31:57 -0800, [EMAIL PROTECTED] (Donald Burrill)
wrote:
> A quibble, and a question (or maybe several, of each), Rich:
[ ... ]
DB: >
> Well, ALL interactions are disordinal, if by "interaction" one means
[ snip, on Don's definition, which Don says is tautological ...]
DB: >
> What most people who
> use "ordinal" and "disordinal" seem to mean is a plot of the cell means
> (or of regression lines), with no adjustment for main effects: so, a
> display that includes the interaction AND the main effects. I take it
> that's what you mean here.
Yes. Just like "most people," I use the definition that has draws a
distinction, instead of the one that does not. Why do you prefer the
one that does not?
DB: >
> Then: a disordinal display -- of what plot?
> (As remarked in a thread a year or two ago, an interaction (displayed as
> a plot of cell means or of regression lines) may appear ordinal from one
> direction and disordinal from the other.)
- I remember someone claimed that.
I remember an example that failed to make the point. I don't
remember a valid example, or that the point was generally accepted.
- I hope this is not a failure of my memory. But if it's my problem,
I hope you will reproduce the illustration, or cite it somewhere.
[ snip, some stuff ]
me: >
> > For instance, in Education:
> > If you group the highest IQ versus lowest,
> > the "regression" for a year or two will be opposite:
> > the highest will learn more new stuff, faster, and get further ahead.
DB: >
> Ah. Now this must surely yield an *ordinal* plot, which implies that
> you would discount this phenomenon as "artifactual" or "pretty trivial"
> (or perhaps both). Do I understand you correctly? Then, which; and
> what (if applicable) is the artifact?
We are talking about an uncontrolled study, where the
sample is characterized on the PRE information. I am describing
a NATURALISTIC outcome, in real life, the followup of no-treatment,
on certain measurements. SO, yes, this result -- which would be
exceptional in some other situations -- is artifactual and trivial,
according to assumptions I am making.
Therefore, the same result might (or might not) be artifactual and
trivial in some other study; no one can tell merely by looking at the
numbers. That is why you need a control group, or you need to
estimate that sort of outcome that a control group might give you.
The only effect that is never potentially artifactual is the crossover
of the means, the Disordinal interaction (as most of us define it).
That one that can't be explained as measurement error (such as,
strong regression owing to poor reliability); scaling (such as,
ceiling effects); or "regression" towards the conditional expected
values (such as, the real-life example I just cited).
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
