Rich Ulrich wrote:
>
> This is a question that I expect that I might answer, except that
> this time, I just can't parse it. In reproducing it here, I have
> inserted different <carriage returns>, to show how I think it is
> phrased.
>
> Is there some typo? think-o?
> Am I framing it wrong?
>
>
> " ... I would like to know, along with several of my colleagues, what
> the advantages are
> for determining the statistical significance of an interaction effect
> when looking at the
> interaction of two or more variables of a population
> as opposed to a sample of the population?"
>
> Sorry, I am missing it -
It's pretty poorly written, but I _think_ that it was intended to be
something like
" ... I, along with several of my colleagues, would like to know what
the advantages are of determining the statistical significance of an
interaction effect when looking at the interaction of two or more
variables, in the case in which we have data for the entire population,
not just a sample."
The answer is that an intact group may (depending on your philosophical
approach) be considered as a quasi-random sample from some hypothetical
population of all might-have-been group members. Thus, if out of ten
chairs of a department, nine had been male, one could say there was
evidence for a pattern of selecting males, whereas if six had been male
one could appeal to statistical arguments and say that even though there
is ironclad evidence that a majority of the *actual* chairs have been
male, there is not evidence of a pattern of discrimination.
A purist would say that, as no random sample has been taken from a
well-defined population, statistical techniques are inappropriate.
However, a purist would say this about almost any real-world study...
-Robert Dawson
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
