Premium
Screening Effects in Multidimensional Contingency Tables
Author(s) -
Brown Morton B.
Publication year - 1976
Publication title -
journal of the royal statistical society: series c (applied statistics)
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.205
H-Index - 72
eISSN - 1467-9876
pISSN - 0035-9254
DOI - 10.2307/2346517
Subject(s) - contingency table , statistics , computer science , econometrics , mathematics
Summary Using the parallelism between the general linear hypothesis and the log‐linear models, we propose that the importance of effects in the log‐linear model for multidimensional contingency tables be studied by computing two test statistics for each effect. These test statistics, called marginal and partial association, indicate the order of magnitude of the change in the tests‐of‐fit when the effect is either entered or deleted from a model. Hence effects may be labelled as definitely needed in the model, definitely not needed, and “uncertain”. The set of models which require further analysis is then limited to those models which include the effects definitely needed and reasonable combinations of the “uncertain” effects.