×

Testing for misspecification in generalized linear mixed models. (English) Zbl 1437.62385

Summary: Generalized linear mixed models have become a frequently used tool for the analysis of non-Gaussian longitudinal data. Estimation is often based on maximum likelihood theory, which assumes that the underlying probability model is correctly specified. Recent research shows that the results obtained from these models are not always robust against departures from the assumptions on which they are based. Therefore, diagnostic tools for the detection of model misspecifications are of the utmost importance. In this paper, we propose 2 diagnostic tests that are based on 2 equivalent representations of the model information matrix. We evaluate the power of both tests using theoretical considerations as well as via simulation. In the simulations, the performance of the new tools is evaluated in many settings of practical relevance, focusing on misspecification of the random-effects structure. In all the scenarios, the results were encouraging, however, the tests also exhibited inflated Type I error rates when the sample size was small or moderate. Importantly, a parametric bootstrap version of the tests seems to overcome this problem, although more research in this direction may be needed. Finally, both tests were also applied to analyze a real case study in psychiatry.

MSC:

62P10 Applications of statistics to biology and medical sciences; meta analysis
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] Examples in which misspecification of a random effects distribution reduces efficiency, and possible remedies, 47, 639-653 (2004) · Zbl 1429.62483 · doi:10.1016/j.csda.2003.12.009
[2] Validation of surrogate markers in multiple randomized clinical trials with repeated measurements: canonical correlation approach, 60, 845-853 (2004) · Zbl 1274.62705 · doi:10.1111/j.0006-341X.2004.00239.x
[3] (2002)
[4] Misspecified maximum likelihood estimates and generalised linear mixed models, 88, 973-985 (2001) · Zbl 0986.62060 · doi:10.1093/biomet/88.4.973
[5] Bootstrap-based critical values for the information matrix test, 61, 395-411 (1994) · doi:10.1016/0304-4076(94)90092-2
[6] Type I and type II error under random-effects misspecification in generalized linear mixed models, 63, 1038-1044 (2007) · Zbl 1274.62822 · doi:10.1111/j.1541-0420.2007.00782.x
[7] The impact of a misspecified random-effects distribution on maximum likelihood estimation in generalized linear mixed models, 27, 3125-3144 (2008) · doi:10.1002/sim.3157
[8] (2007)
[9] (2005)
[10] A note on the uniformly most powerful tests in the presence of nuisance parameters, 44, 141-145 (1992) · Zbl 0763.62009 · doi:10.1007/BF00048676
[11] (1993)
[12] The size bias of White’s information matrix test, 24, 63-67 (1987) · Zbl 1328.62153 · doi:10.1016/0165-1765(87)90182-0
[13] A diagnostic test for the mixing distribution in a generalized linear mixed model, 93, 1003-1010 (2006) · Zbl 1436.62364 · doi:10.1093/biomet/93.4.1003
[14] The effect of misspecifying the random-effects distribution in linear mixed models for longitudinal data, 23, 541-556 (1997) · Zbl 0900.62374 · doi:10.1016/S0167-9473(96)00047-3
[15] A simulation-based goodness-of-fit test for random effects in generalized linear mixed models, 33, 721-731 (2006) · Zbl 1164.62348 · doi:10.1111/sjos.2006.33.issue-4
[16] Maximum likelihood estimation of misspecified models, 50, 1-25 (1982) · Zbl 0478.62088 · doi:10.2307/1912526
[17] Likelihood-ratio DIF testing: effects of nonnormality, 32, 511-526 (2008) · doi:10.1177/0146621607310402
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.