Approximate score-based testing with application to multivariate trait association analysis

Alzheimer's Disease Neuroimaging Initiative

Research output: Contribution to journalArticlepeer-review

6 Scopus citations

Abstract

For genome-wide association studies and DNA sequencing studies, several powerful score-based tests, such as kernel machine regression and sum of powered score tests, have been proposed in the last few years. However, extensions of these score-based tests to more complex models, such as mixed-effects models for analysis of multiple and correlated traits, have been hindered by the unavailability of the score vector, due to either no output from statistical software or no closed-form solution at all. We propose a simple and general method to asymptotically approximate the score vector based on an asymptotically normal and consistent estimate of a parameter vector to be tested and its (consistent) covariance matrix. The proposed method is applicable to both maximum-likelihood estimation and estimating function-based approaches. We use the derived approximate score vector to extend several score-based tests to mixed-effects models. We demonstrate the feasibility and possible power gains of these tests in association analysis of multiple and correlated quantitative or binary traits with both real and simulated data. The proposed method is easy to implement with a wide applicability.

Original languageEnglish (US)
Pages (from-to)469-479
Number of pages11
JournalGenetic epidemiology
Volume39
Issue number6
DOIs
StatePublished - Sep 1 2015

Keywords

  • GWAS
  • Kernel machine regression
  • Mixed-effects models
  • Multiple traits
  • SNP
  • Sum of powered score (SPU) tests

Fingerprint Dive into the research topics of 'Approximate score-based testing with application to multivariate trait association analysis'. Together they form a unique fingerprint.

Cite this