Abstract
The generalized linear mixed model (GLMM) extends classical regression analysis to non-normal, correlated response data. Because inference for GLMMs can be computationally difficult, simplifying distributional assumptions are often made. We focus on the robustness of estimators when a main component of the model, the random effects distribution, is misspecified. Results for the maximum likelihood estimators of the Poisson inverse Gaussian model are presented.
Similar content being viewed by others
References
Abramowitz M, Stegun I (eds) (1972) Handbook of mathematical functions. Dover, New York
Bickel PJ (1984) Robust regression based on infinitesimal neighbourhoods. Ann Stat 12:1349–1368
Bickel PJ, Doksum KA (2001) Mathematical statistics: basic ideas and selected topics. Prentice Hall, Upper Saddle River
de Bruijn NG (1953) The difference-differential equation \(f^\prime (x) = e^\alpha x + \beta f(x-1)\), i, ii. Indag Math 15(449–458):459–464
Dean CB, Nielsen JD (2007) Generalized linear mixed models: a review and some extensions. Lifetime Data Anal 13(4):497–512
Dean C, Lawless JF, Willmot GE (1989) A mixed Poisson-inverse-Gaussian regression model. Canad J Stat 17:171–181
Feller W (1966) An introduction to probability theory and its applications, vol 2. Wiley, New York
Gustafson P (1996) The effect of mixing-distribution misspecification in conjugate mixture models. Canad J Stat 24:307–318
Hampel FR (1974) The influence curve and its role in robust estimation. J Am Stat Assoc 69:383–393
Hampel FR, Ronchetti EM, Rousseeuw PJ, Stahel WA (1986) Robust statistics: the approach based on influence functions. Wiley, New York
Heagerty PJ, Zeger SL (2000) Marginalized multilevel models and likelihood inference (with comments and a rejoinder by the authors). Stat Sci 15(1):1–26
Heckman J, Singer B (1984) A method for minimizing the impact of distributional assumptions in econometric models for duration data. Econom J Econom Soc 52(2):271–320
Hilbe JM (2014) Modeling count data. Cambridge University Press, New York
Hougaard P, Lee ML, Whitmore GA (1997) Analysis of overdispersed count data by mixtures of Poisson variables and Poisson processes. Biometrics 53:1225–1238
Huber PJ (1981) Robust statistics. Wiley, New York
Karlis D, Xekalaki E (2005) Mixed Poisson distributions. Int Stat Rev 73(1):35–58
Litière S, Alonso A, Molenberghs G (2008) The impact of a misspecified random-effects distribution on the estimation and the performance of inferential procedures in generalized linear mixed models. Stat Med 27(16):3125–3144
McCulloch CE, Neuhaus JM (2013) Misspecifying the shape of a random effects distribution: why getting it wrong may not matter. Biostatistics 14:477–490
McCulloch CE, Searle SR, Neuhaus JM (2008) Generalized, linear and mixed models. Wiley, New York
Neuhaus JM, Hauck WW, Kalbfleisch JD (1992) The effects of mixture distribution misspecification when fitting mixed-effects logistic models. Biometrika 79:755–762
Ong SH (1998) A note on the mixed Poisson formulation of the Poisson-inverse Gaussian distribution. Commun Stat Simul Comput 27(1):67–78
R Core Team. R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, 2017. https://www.R-project.org
Rieder H (1994) Robust Asymptot Stat. Springer, New York
Seshadri V (1993) The inverse Gaussian distribution: a case study in exponential families. Clarendon, Oxford
Seshadri V (1999) The inverse Gaussian distribution: statistical theory and applications. Springer, New York
Shaban SA (1981) Computation of the Poisson-inverse Gaussian distribution. Commun Stat Theory Methods 10(14):1389–1399
Shoukri MM, Asyali MH, VanDorp R, Kelton DF (2004) The Poisson inverse Gaussian regression model in the analysis of clustered counts data. J Data Sci 2:17–32
Stasinopoulos MD, Rigby RA, Heller GZ, Voudouris V, De Bastiani F (2017) Flexible regression and smoothing: using GAMLSS in R. Chapman and Hall/CRC, London
Teugels JL, Willmot G (1987) Approximations for stop-loss premiums. Insur Math Econ 6:195–202
Van de Ven R, Weber NC (1995) Log-linear models for mean and dispersion in mixed Poisson regression models. Aust J Stat 37(2):205–216
Verbeke G, Molenberghs G (2013) The gradient function as an exploratory goodness-of-fit assessment of the random-effects distribution in mixed models. Biostatistics 14:477–490
Weems KS, Smith PJ (2004) On robustness of maximum likelihood estimates for Poisson-lognormal models. Stat Probab Lett 66:189–196
Willmot GE (1990) Asymptotic tail behaviour of Poisson mixtures with applications. Adv Appl Probab 22:147–159
Zha L, Lord D, Zou Y (2016) The Poisson inverse Gaussian (PIG) generalized linear regression model for analyzing motor vehicle crash data. J Transp Saf Secur 8(1):18–35
Acknowledgements
The authors thank the Editor, reviewers, Dr. Dennis Boos (North Carolina State University) and Dr. Kimberly F. Sellers (Georgetown University) for helpful comments and suggestions that significantly improved this paper.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
On behalf of all authors, the corresponding author states that there is no conflict of interest.
Additional information
This work was supported by National Science Foundation Grant #1700235. Additional support was provided by a grant from North Carolina Central university.
Appendix: Upper and lower bounds for Poisson inverse Gaussian probability ratios
Appendix: Upper and lower bounds for Poisson inverse Gaussian probability ratios
Below we provide proofs to Lemmas 2 and 3, in which we find an upper bound and a lower bound for Poisson inverse Gaussian probability ratios.
Lemma 2
Let \(\nu =+\sqrt{\tau ^{-1}(\tau ^{-1} + 2\mu )}\). Then for \(y>0,\)
Proof
Equation (7) gives us the following relationship:
From Abramowitz and Stegun (1972), we have that for \(k=0,\pm \,1,\pm \,2,\ldots \)
where \({C(\nu )}={\sqrt{\pi /(2\nu )} \exp (-\nu )}\). Therefore, we may write
Substituting in the numerator of (13) we find that
where the last line uses the following inequality:
\(\square \)
Lemma 3
Let \(\nu =+\,\sqrt{\tau ^{-1}(\tau ^{-1} + 2\mu )}\). Then for \(y>0,\)
Proof
Recall from Eq. (13) that
Using (14) with \({C(\nu )}={\sqrt{\pi /(2\nu )} \exp (-\nu )}\), we may write
Therefore, substituting into (16), we have the following lower bound:
where we use the elementary inequality \(1/(1-x) \ge 1+x\). Notice that
Therefore, (17) becomes
by substitution. \(\square \)
Rights and permissions
About this article
Cite this article
Weems, K.S., Smith, P.J. Assessing the robustness of estimators when fitting Poisson inverse Gaussian models. Metrika 81, 985–1004 (2018). https://doi.org/10.1007/s00184-018-0664-1
Received:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00184-018-0664-1