On 12 Nov 2007 at 13:03, Ken Steele wrote: > What was the N for the groups? > > One justification for using the covariates might be that the N > was very low and that variability within the groups needed > further reduction.
N = 69 for the intervention group; 181 for control group (not
particularly low, I'd think, at least for a randomized experimental
study).
But there is something that's curious about these groups. For the
reported comparisons, there were actually two different controls. One was
an "attention-control" ("no treatment sensitization/contact control"; n =
123); the other was just a "control" ("no-treatment, no-contact control";
n = 58). They didn't give details, but I assume the attention-control was
a kind of minimal placebo group. But after going to the trouble of
creating two kinds of control group, they didn't report any comparisons
using them, except for comparing one with the other. Finding no
difference, they used that as justification for combining them. All
comparisons with the intervention group were made using the combined
group.
That's a problem because their control group then is a strange hybrid,
neither purely no-treatment nor attention-placebo. I suspect they did it
because they weren't getting much joy from their comparisons with each
control group separately. But it does limit whart they can conclude about
their intervention (not that they felt so constrained, of course).
The disparity in the size of the groups is also strange. They don't say
why, and it doesn't seem to be due to drop-outs (which they don't report
either). Possibly it was because each intervention procedure took a lot
of time to administer, and each control, very little, so they could
handle many more controls than intervention subjects.
Back to my posted question on the legitimacy of logistic regression in
this situation. On the basis of the advice I've received here (and
thanks for the helpful responses), I'm now persuaded that it's a
defensible procedure, and even a good idea, because it buys them greater
power at no cost. I should be happy and not worry about it. Right?
Stephen
-----------------------------------------------------------------
Stephen L. Black, Ph.D.
Professor of Psychology, Emeritus
Bishop's University e-mail: [EMAIL PROTECTED]
2600 College St.
Sherbrooke QC J1M 1Z7
Canada
-----------------------------------------------------------------------
---
WPM$0BEF.PM$
Description: Mail message body
