Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.
|
|
|
|
Re: How defend nonrandom samples?
Posted:
Sep 22, 1996 1:45 AM
|
|
YES, Warren --
I'll join you on the soap box. I'm with you on the importance of dealing
with LURKING VARIABLES (as Moore & McCabe (2nd Edition)indicate). In
observational studies, the capability of dealing with LURKING VARIABLES is
a "selling point" for many students when they realize that it may be
feasible to CONTROL FOR THE UNCONTROLLABLE! So one of my favorite
expressions is: IF YOUR CAN'T CONTROL IT, MEASURE IT.
Of course you can't measure all of the LURKING VARIABLES. But when
students realize that they might MATHEMATICALLY CONTROL for RAINFALL if
they can measure it, then they might just become aware of the possibility
that further study of statistics is useful and interesting.
IMHO, it is essential that students are made to really believe that they
can do something useful AFTER COMPLETING THEIR FIRST STATISTICS COURSE
THAT THEY COULD NOT DO BEFORE! A possible reason that many one-course
students say "Uuugggg" when Statistics is mentioned, is that they can't do
much useful research after they have taken one course.
-- Joe
***********************************************************************
* Joe Ward 167 East Arrowhead Dr. *
* Health Careers High School San Antonio, TX 78228-2402 *
* Phone: 210-433-6575 *
* joeward@tenet.edu *
***********************************************************************
On Sat, 21 Sep 1996, Warren wrote:
> Date: Sat, 21 Sep 1996 23:58:45 -0400
> From: Warren <wlmay@umsmed.edu>
> To: Multiple recipients of list <edstat-l@jse.stat.ncsu.edu>
> Subject: Re: How defend nonrandom samples?
>
> rbeldin@world.std.com (Richard A Beldin) wrote:
>
> >In reality, few research programs can employ such techniques as random
> >sampling, double-blind treatment assignments, and such. That is why so
> >much junk appears in the popular literature. We have given up trying to
> >shout down the purveyors of simulated surveys and ersatz experiments.
> >
> Seems to me Richard brings up a very good point here. I'm working in a
> Med Ctr. environment where many of the studies are observational. All of
> the neat little study designs, balanced designs, normality assumptions,
> uncorrelated error terms, complete data, lack of interaction effects,
> etc., are class notes. The data are often very messy. Ethical
> considerations prohibit randomization in some instances (someone just
> left my office a few minutes ago with that very problem). The patient
> has a right to proper treatment and placebo controlled studies aren't
> always possible. In a great many studies, the sample is taken for
> convenience...it has to be to be timely. We can't wait to take, say,
> every 10th person. Even though research is a part of the MD's
> responsibility, that's not their sole purpose.
>
> So, does the data not have value just because it is not a random sample?
> If 1000 people eat Aunt Jane's potato salad at a picnic, 200 report to
> the hospital after hearing the 10 PM news, 100 of these are ill and 10
> die we can pretty much conclude that something was rotten. Do we need
> all 1000 to say that? The 200 aren't a random sample, but they do tell
> us the story.
>
> Stepping down from soapbox now,
> Warren.
>
> NEXT!
>
>
|
|
|
|
