Premium
On Significance Testing in Comparative Epidemiologic Research
Author(s) -
Nurminen M.,
Mutanen P.,
Järvinen E.
Publication year - 1982
Publication title -
biometrical journal
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.108
H-Index - 63
eISSN - 1521-4036
pISSN - 0323-3847
DOI - 10.1002/bimj.4710240806
Subject(s) - significance testing , statistical hypothesis testing , econometrics , psychology , statistics , mathematics
The choice of the referent entity in comparative epidemiologic studies is, of course, crucial to the attainment of a valid and sharp contrast. A commonplace procedure in large sample significance testing for (multiple) fourfold (and extended) tables is to base the estimation of the expected value (and/or the variance) on the combined study and comparison experience. This procedure is, however, deplorable in that in interim stages of the study it may hide the underlying difference in rates. The notion that the probability of the realization of the observed number of exposed cases be evaluated unconditionally assuming a binomial model with the Bernoulli parameter estimated solely by the rate in the referent series is considered. This outlook is carried over to bilateral testing. A chi square criterion accommodating this viewpoint could be used to canvass the accrued data for deciding on the prospects of continuing the data collection. A test employing exact variance estimates is derived and various methods for computing an ‘exact’ significance probability are developed. The procedure is exemplified by artificial data and accompanied by a discussion of its applicability.