Proper test result to choose

Hi,

This may be an old question.

I conduct two tests (logistic regression and Fisher's exact test) to
analyze my data. One predictor is signicant in logistic regression
(p=0.03) and not significant in Fisher's exact test (p=0.10). As no
pre-determined method is specified, which is the proper method I should
use.

In addition, what makes the difference? Is the logistics regression
more "powerful"?

Thanks,

Wilson

.

Relevant Pages

  • Re: Proper test result to choose
    ... I conduct two tests (logistic regression and Fisher's exact test) to ... The "likelihood ratio" test for the logistic regression (i.e., the chi-square test on the change in -2 log likelihood) is equivalent to the "likelihood ratio chi-square" you would get on the 2x2 table. ... One guideline states that you ought to have at least 10 events per predictor. ...
    (sci.stat.edu)
  • Re: Proper test result to choose
    ... > I conduct two tests (logistic regression and Fisher's exact test) to ... One predictor is signicant in logistic regression ... Those two tests are not considered as competitors. ... correction, which is practically identical; ...
    (sci.stat.edu)