Re: Reasonable use of "significance" testing


Bruce Weaver wrote:
Reef Fish wrote:
Bruce Weaver wrote:
Herman Rubin wrote:
I know of one person who uses what I consider a somewhat
reasonable use of significance testing.

He formulates a model, and then tests until it is rejected
at the 5% level. Then he looks at the fit, and if it is
good enough, he accepts the model as an approximation.

This is not optimal; one should not take too many observations,
as the expected number of observations if the null hypothesis
is true is infinite, and if it is almost true is large. But
at least this is a good way to do research.

What is his area of research?

--
Bruce Weaver
bweaver@xxxxxxxxxxxx
www.angelfire.com/wv/bwhomedir

Excellent question.

-- Bob.



The response from Professor Rubin was "psychology".

Figures.

Thanks.

I learned EVERYTHING I know about psychology from Dave Barry,
who taught me that:

PSYCHOLOGY: This involves talking about rats and dreams.
Psychologists are obsessed with rats and dreams. I once
spent an entire semester training a rat to punch little buttons
in a certain sequence, then training my roommate to do the
same thing. The rat learned much faster. My roommate is
now a doctor. If you like rats or dreams, and above all if you
dream about rats, you should major in psychology.

That's one of the reasons I majored in statistics. :)

-- Bob.

.

Quantcast