Donald Burrill <[EMAIL PROTECTED]> wrote in message
[EMAIL PROTECTED]">news:[EMAIL PROTECTED]...
> Clarification(s), please:
>
> On Wed, 23 Aug 2000, Ken Reed wrote:
>
> > I'm trying to test whether a variable measures a group-level property,
> > and so I'm looking for an analog to eta-squared, intra-class correlation
> > etc for nominal or ordinal data.
>
> Do you have a particular group-level property in mind, that is measured
> by some variable(s) other than the one you're trying to test? Or are you
> trying to infer the existence of some such property from the behavior of
> this variable alone (or, perhaps, in concert with others)?
The latter.
>
> > I have data comprising 2000 workplaces, within samples of individuals
> > drawn from each (n=20,000).
>
> Random samples, or convenience samples? Ten from each workplace, or
> variable (and if variable, why?)? Do the workplaces assort themselves
> into categories, or are you looking at something like a 1-way ANOVA with
> 2000 groups?
I should have said 'with samples of individuals' not 'within samples of
individuals'.
The workplaces are randomly sampled. Selection of the individuals was based
on a random sample design, but there are known biases.
You've got it -- I'm looking for something analogous to a 1-way ANOVA with
2000 groups, but with a nominal dependent variable.
>
> > One variable has 4 categories (agree-neutral-disagree, don't know).
>
> Are you trying to say that you have one such variable, and your other
> variables are otherwise described; or that you have a number of such
> variables and you want something like an item analysis of them all; or
> that you have a number of such variables that you intend to combine in
> some unspecified way to produce the variable you want to test; or ...?
My question is about just the one variable.
>
> By the way, this variable is not one variable, it is two: (1) degree of
> agreement with whatever, and (2) whether the respondent has an opinion
> about it. If you have a bunch of variables like this, what you can do
> with them depends partly on how much missing data (= "don't know"
> responses) you have.
"don't know" is a meaningful response in this case -- it is not 'missing'
data.
>
> > 1. How can I estimate how much of the total variability derives from
> > between groups (workplaces) and within groups?
>
> "Total variability" of one of these 4-category variables, or of a total
> score derived from a bunch of them, or of the bunch of them considered as
> a multivariate whole?
One variable -- the problems analagous to one-way ANOVA, but with a nominal
variable.
>
> > 2. Is there a rule-of-thumb for what would be evidence of strong
> > within-group agreement?
>
> Rules of thumb exist only to help one avoid having to think hard about
> some situation or problem. As such they are invariably heavily
> dependent on contexts, about which we have very little information.
Perhaps. But they also signal conventions.
>
> ------------------------------------------------------------------------
> Donald F. Burrill [EMAIL PROTECTED]
> 348 Hyde Hall, Plymouth State College, [EMAIL PROTECTED]
> MSC #29, Plymouth, NH 03264 603-535-2597
> 184 Nashua Road, Bedford, NH 03110 603-471-7128
>
>
>
> =================================================================
> Instructions for joining and leaving this list and remarks about
> the problem of INAPPROPRIATE MESSAGES are available at
> http://jse.stat.ncsu.edu/
> =================================================================
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
