Premium
The longitudinal nonparametric test as a new tool to explore gene‐gene and gene‐time effects in cohorts
Author(s) -
Malzahn D.,
Schillert A.,
Müller M.,
Bickeböller H.
Publication year - 2010
Publication title -
genetic epidemiology
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 1.301
H-Index - 98
eISSN - 1098-2272
pISSN - 0741-0395
DOI - 10.1002/gepi.20500
Subject(s) - nonparametric statistics , trait , normality , parametric statistics , statistics , quantitative trait locus , type i and type ii errors , biology , genetics , normal distribution , econometrics , mathematics , gene , computer science , programming language
Current approaches for analysis of longitudinal genetic epidemiological data of quantitative traits are typically restricted to normality assumptions of the trait. We introduce the longitudinal nonparametric test (LNPT) for cohorts with quantitative follow‐up data to test for overall main effects of genes and for gene‐gene and gene‐time interactions. The LNPT is a rank procedure and does not depend on normality assumptions of the trait. We demonstrate by simulations that the LNPT is powerful, keeps the type‐1 error level, and has very good small sample size behavior. For phenotypes with normal residuals, loss of power compared to parametric approaches (linear mixed models) was small for the quite general scenarios, which we simulated. For phenotypes with non‐normal residuals, gain in power by the LNPT can be substantial. In contrast to parametric approaches, the LNPT is invariant with respect to monotone transformations of the trait. It is mathematically valid for arbitrary trait distribution. Genet. Epidemiol . 34: 469‒478, 2010. © 2010 Wiley‐Liss, Inc.