[R-sig-Epi] could someone tell me how to implement a multiple comparison test for proportions in a 2xc crosstabulation
Kevin Viel
kviel at darwin.sfbr.org
Wed Jun 14 18:03:04 CEST 2006
xingwang ye wrote:
> I wanna to do multiple comparison test for proportions, could someone
> tell me how in R, thank you!
>
> _______________________________________________
> R-sig-Epi at stat.math.ethz.ch mailing list
> https://stat.ethz.ch/mailman/listinfo/r-sig-epi
Typically, we formulate a single hypothesis and test it:
The frequency of exposure in the cases is higher than the frequency of
exposure in controls.
Sometimes we alter this a bit:
The frequency of exposure in the cases is higher than the frequency of
exposure in:
(a) Control subgroup A
(b) Control subgroup B
.
.
.
(n) Control subgroup N
If you have enough groups, then some of the comparisons might appear
suggestive by chance alone (1 in 20 at alpha = 0.05).
One might attempt to account for this via corrections for multiple
comparisons. In doing this, one increases the stringency of the p-value
(or other evidence). A conservative approach is the Bonferroni
correction: the critical value for the p-value becomes the nominal value
divided by the number of test. For instance, at alpha 0.05 and ten
tests the p-value for "significance" becomes 0.005.
You may have something else in mind. If you write back to the list and
describe the investigation (hypotheses), then we might be able to assist
you better.
Good luck,
Kevin
----------------------------------------------------------------------
Kevin Viel
Department of Genetics e-mail: kviel at darwin.sfbr.org
Southwest Foundation phone: (210)258-9884
P.O. Box 760549 fax: (210)258-9444
San Antonio, TX 78245-0549
Kevin Viel
PhD Candidate
Department of Epidemiology
Rollins School of Public Health
Emory University
Atlanta, GA 30322
More information about the R-sig-Epi
mailing list