In article <[EMAIL PROTECTED]>,
Rich Ulrich <[EMAIL PROTECTED]> wrote:
>On Fri, 03 Mar 2000 18:19:18 +1600, szabos
><[EMAIL PROTECTED]> wrote:
>> In my case-control study, I have two separate contingency
>> tables for gender. My chi square analysis showed
>> statistically insignificant results for both genders. Yet
>> when I pooled the two contingency tables into one large
>> table, gender unbiased, I got statistically significant
>> results...why is this?
>> My only reason would be a larger sample size?
>> Any others?
>Where (x,y) is a row, and {(,), (,)} is a table,
>consider
> {(1,4), (4,16)} and
> {(16,4),(4,1)} -- each with chisquared=0,
>and which give a total of {(17,8),(8,17)}, chisquared= 6.48.
>Some topics to read about are "Simpson's paradox" or the ecological
>fallacy, and statistical confounding.
This is a major problem in many situations. It is
exascerbated by the general inability to take into
account ALL factors, to carry out studies in which
key factors are uncontrolled, and often the results
obtained cause policies to be introduced and laws
passed which only make the real problem worse.
The results can be spurious dependency, or in some
cases, spurious independence. We can have the model
y = a_0 + a_1*x_1 + a_2*x_2,
where a_1 and a_2 are both definitely not 0, but
where the correlation between x_2 and y is close
enough to 0 that the influence of x_2 and y can
only be detected when x_1 is included. In many
cases, the model is more important than even a
lot of observations.
--
This address is for information only. I do not claim that these views
are those of the Statistics Department or of Purdue University.
Herman Rubin, Dept. of Statistics, Purdue Univ., West Lafayette IN47907-1399
[EMAIL PROTECTED] Phone: (765)494-6054 FAX: (765)494-0558
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
