Saddlepoint tests for accurate and robust inference on overdispersed count data. (English) Zbl 1466.62013

Summary: Inference on regression coefficients when the response variable consists of overdispersed counts is traditionally based on Wald, score and likelihood ratio tests. As the accuracy of the \(p\)-values of such tests becomes questionable in small samples, three recently developed tests are adapted to the negative binomial regression model. The non-trivial computational aspects involved in their implementation, some of which remained obscure in the literature until now, are detailed for general \(M\)-estimators. Under regularity conditions, these tests feature a relative error property with respect to the asymptotic chi-squared distribution, thus yielding highly accurate \(p\)-values even in small samples. Extensive simulations show how these new tests outperform the traditional ones in terms of actual level with comparable power. Moreover, inference based on robust (bounded influence) versions of these tests remains reliable when the sample does not entirely conform to the model assumptions. The use of these procedures is illustrated with data coming from a recent randomized controlled trial, with a sample size of 52 observations. An R package implementing all tests is readily available.


62-08 Computational methods for problems pertaining to statistics


robNB; R
Full Text: DOI


[1] Aeberhard, W. H.; Cantoni, E.; Heritier, S., Robust inference in the negative binomial regression model with an application to falls data, Biometrics, 70, 4, 920-931, (2014) · Zbl 1393.62011
[2] Almudevar, A.; Field, C.; Robinson, J., The density of multivariate M-estimates, Ann. Statist., 28, 1, 275-297, (2000) · Zbl 1106.62335
[3] Beran, R., Prepivoting to reduce level error of confidence sets, Biometrika, 74, 3, 457-468, (1987) · Zbl 0663.62045
[4] Brown, B. W.; Newey, W. K., Efficient semiparametric estimation of expectations, Econometrica, 66, 2, 453-464, (1998) · Zbl 1008.62571
[5] Camponovo, L., Otsu, T., 2015. Relative error refinement in nonparametric likelihood, working paper. · Zbl 1407.62156
[6] Canning, C. G.; Sherrington, C.; Lord, S. R.; Close, J. C.T.; Heritier, S.; Heller, G. Z.; Howard, K.; Allen, N. E.; Latt, M. D.; Murray, S. M.; O’Rourke, S. D.; Paul, S. S.; Song, J.; Fung, V. S.C., Exercise for falls prevention in parkinson disease, Neurology, 84, 3, 304-312, (2015)
[7] Cordeiro, G. M., Improved likelihood ratio statistics for generalized linear models, J. R. Stat. Soc. Ser. B, 45, 3, 404-413, (1983) · Zbl 0529.62024
[8] Cordeiro, G. M.; Ferrari, S. L.P., A modified score test statistic having chi-squared distribution to order \(n^{- 1}\), Biometrika, 78, 3, 573-582, (1991) · Zbl 1192.62053
[9] Czellar, V.; Ronchetti, E. M., Accurate and robust tests for indirect inference, Biometrika, 97, 3, 621-630, (2010) · Zbl 1195.62047
[10] Dean, C. B., Modified pseudo-likelihood estimator of the overdispersion parameter in Poisson mixture models, J. Appl. Stat., 21, 6, 523-532, (1994)
[11] DiCiccio, T.; Hall, P.; Romano, J., Empirical likelihood is bartlett-correctable, Ann. Statist., 19, 2, 1053-1061, (1991) · Zbl 0725.62042
[12] Field, C.; Robinson, J.; Ronchetti, E. M., Saddlepoint approximations for multivariate M-estimates with applications to bootstrap accuracy, Ann. Inst. Statist. Math., 60, 1, 205-224, (2008) · Zbl 1184.62017
[13] Hall, P.; Martin, M. A., On bootstrap resampling and iteration, Biometrika, 75, 4, 661-671, (1988) · Zbl 0659.62053
[14] Hampel, F. R.; Ronchetti, E. M.; Rousseeuw, P. J.; Stahel, W. A., Robust statistics: the approach based on influence functions, (1986), Wiley New York, NY · Zbl 0593.62027
[15] Heritier, S.; Ronchetti, E. M., Robust bounded-influence tests in general parametric models, J. Amer. Statist. Assoc., 89, 427, 897-904, (1994) · Zbl 0804.62037
[16] Hinde, J.; Demétrio, C. G.B., Overdispersion: models and estimation, Comput. Statist. Data Anal., 27, 2, 151-170, (1998) · Zbl 1042.62578
[17] Imbens, G. W.; Spady, R. H.; Johnson, P., Information theoretic approaches to inference in moment condition models, Econometrica, 66, 2, 333-357, (1998) · Zbl 1055.62512
[18] Kitamura, Y.; Stutzer, M., An information-theoretic alternative to generalized method of moments estimation, Econometrica, 65, 4, 861-874, (1997) · Zbl 0894.62011
[19] Lô, S. N.; Ronchetti, E. M., Robust and accurate inference for generalized linear models, J. Multivariate Anal., 100, 9, 2126-2136, (2009) · Zbl 1222.62093
[20] Lô, S. N.; Ronchetti, E. M., Robust small sample accurate inference in moment condition models, Comput. Statist. Data Anal., 56, 11, 3182-3197, (2012) · Zbl 1255.62370
[21] Lunardon, N.; Ronchetti, E. M., Composite likelihood inference by nonparametric saddlepoint tests, Comput. Statist. Data Anal., 79, 80-90, (2014)
[22] Ma, Y.; Ronchetti, E. M., Saddlepoint test in measurement error models, J. Amer. Statist. Assoc., 106, 493, 147-156, (2011) · Zbl 1396.62108
[23] R Core Team. 2015. R: A Language and Environment for Statistical Computing. R Foundation for Statistical Computing, Vienna, Austria.
[24] Robinson, J.; Ronchetti, E. M.; Young, G. A., Saddlepoint approximations and tests based on multivariate M-estimates, Ann. Statist., 31, 4, 1154-1169, (2003) · Zbl 1056.62023
[25] Ronchetti, E. M.; Ventura, L., Between stability and higher-order asymptotics, Stat. Comput., 11, 1, 67-73, (2001)
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. It attempts to reflect the references listed in the original paper as accurately as possible without claiming the completeness or perfect precision of the matching.