%0 Journal Article
%T A Nonparametric Test of Interaction in the General Two-Way Layout
%A John £¿HRVIK
%J Turkiye Klinikleri Journal of Biostatistics
%D 2010
%I Turkiye Klinikleri
%X The two-way layout is frequently occurring, e.g. blocking is used to reduce the between subject variability when comparing treatments or many medical centers are included in a clinical trial to recruit a sufficient number of patients. In epidemiological studies, it is common to study the interactions between genetic and environmental factors. This paper is concerned with the statistical analysis of data arising in these situations when assumptions like normality do not necessarily apply. The main objective of this paper is to propose a test for interactions for continuous data based on joint ranking of all observations after iteratively eliminating the two main effects. The validity of the significance levels of the test when using a finite sample version of the asymptotic distribution of the test statistic is manifested and the power against different alternatives illustrated by extensive simulation experiments. The proposed test is compared with competitors on published data sets. Data from the Survey of Adolescent Life in Vestmanland (SALVe) project are analyzed both with our test and Brunner and Puri's proposal. The test shows good asymptotic and small sample properties. It can be used in the general two-way layout and multiple comparisons can be performed in a straightforward way. The analyses of the SALVe project show that our proposal can be useful in such studies. It has the potential to pick-up interactions hidden in the noise of gross errors when using standard ANOVA. The test may become a valuable tool and alternative to other proposals in exploring interactions.
%K Epistasis
%K genetic
%K statistics
%K nonparametric
%K computer simulation
%K research design
%U http://www.turkiyeklinikleri.com/pdf/?pdf=15fe6c567c6e8026e57e5bd88b17609e