Reconciling Fisher and Neyman
- 168 Downloads
The major themes of the previous chapters are
The randomized controlled clinical trial is a complex scientific experiment that is expensive to run and that yields a rich lode of data.
Probability calculations can be made from data, but there is no satisfactory way of relating those probabilities to real life.
Significance testing, as developed by Fisher, is a relatively vague tool, in which probabilities are calculated as a means of rejecting null hypotheses, which are postulated as straw-men.
Significance tests must be directed against a well-defined class of alternative hypotheses that represent what is really expected to happen, and the narrower the class of alternatives, the more powerful the test.
Confidence intervals can be calculated through the use of the algoroithms of significance test, but they can be applied to randomly chosen subsets of the data without invoking the problems associated with significance test applied to subsets of patients.
KeywordsSignificance Test Percent Coverage Numerical Answer Construct Confidence Interval Medical Question
Unable to display preview. Download preview PDF.