Date: Wed, 5 Sep 2007 20:05:42 -0700
Reply-To: David L Cassell <davidlcassell@MSN.COM>
Sender: "SAS(r) Discussion" <SAS-L@LISTSERV.UGA.EDU>
From: David L Cassell <davidlcassell@MSN.COM>
Subject: Re: Statistics question: What does this reviewer want?
Content-Type: text/plain; format=flowed
tree.frog2@HOTMAIL.COM sagely replied:
>On Sep 4, 11:01 am, peterflomconsult...@mindspring.com (Peter Flom)
> > I can't provide too much specifics, for reasons of confidentiality, but
>a colleague submitted an article and got back a review that puzzled him.
>He asked me for help, and it puzzles me, too.
> > The dependent variable is dichotomous, the presence or absence of a
>condition and it's a case-control study . The independent variables are
>varied: Some continuous, some categorical. Some are demographic, but the
>key ones are clinical.
> > Originally, the author compared the groups in a series of MannWhitney U
>tests, and Fisher-Freeman-Hamilton tests, one independent variable at a
>time. The reviewer asked for regression (that part's easy) and for methods
>that would test proposed causal pathways.... The reviewer was very
>imprecise. We're not sure what he/she wants, and he/she may not be, either.
> > I plan to enter variables in particular orders into logistic regression,
>but what else could he/she want?
> > If you need more details, write me off list and I'll supply some.
> > Thanks
> > Peter
>Structural Equation Modeling?
That's my guess, as well. But it may not be feasible to perform SEM
on your data, and it may not be reasonable to attempt it given your
project goals, and SEM may not return useful information.
In my opinion, causality is a matter for scientists and not statisticians.
The stats show the relationships. The science behind the data is
where you get the causality.
David L. Cassell
3115 NW Norwood Pl.
Corvallis OR 97330
Test your celebrity IQ. Play Red Carpet Reveal and earn great prizes!