Document Server@UHasselt >
Research >
Research publications >

Please use this identifier to cite or link to this item: http://hdl.handle.net/1942/15396

Title: The gradient function as an exploratory goodness-of-fit assessment of the random-effects distribution in mixed models
Authors: Verbeke, Geert
Molenberghs, Geert
Issue Date: 2013
Citation: BIOSTATISTICS, 14 (3), p. 477-490
Abstract: Inference in mixed models is often based on the marginal distribution obtained from integrating out random effects over a pre-specified, often parametric, distribution. In this paper, we present the so-called gradient function as a simple graphical exploratory diagnostic tool to assess whether the assumed random-effects distribution produces an adequate fit to the data, in terms of marginal likelihood. The method does not require any calculations in addition to the computations needed to fit the model, and can be applied to a wide range of mixed models (linear, generalized linear, non-linear), with univariate as well as multivariate random effects, as long as the distribution for the outcomes conditional on the random effects is correctly specified. In case of model misspecification, the gradient function gives an important, albeit informal, indication on how the model can be improved in terms of random-effects distribution. The diagnostic value of the gradient function is extensively illustrated using some simulated examples, as well as in the analysis of a real longitudinal study with binary outcome values.
Notes: Katholieke Univ Leuven, Interuniv Inst Biostat & Stat Bioinformat, B-3000 Louvain, Belgium. Univ Hasselt, Interuniv Inst Biostat & Stat Bioinformat, B-3590 Diepenbeek, Belgium.
URI: http://hdl.handle.net/1942/15396
DOI: 10.1093/biostatistics/kxs059
ISI #: 000320433000006
ISSN: 1465-4644
Category: A1
Type: Journal Contribution
Validation: ecoom, 2014
Appears in Collections: Research publications

Files in This Item:

Description SizeFormat
Published version472.77 kBAdobe PDF

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.