Abstract
This paper mainly discusses two issues about asymptotic expansions for the distributions of \(\chi ^2\)-type test statistics. First, it is shown that the generalized empirical likelihood ratio, Wald-type, and score-type test statistics for a subvector hypothesis in the possibly over-identified moment restrictions are, in general, not Bartlett-correctable, except for the empirical likelihood ratio test statistic. Second, starting with the classical likelihood or the modern generalized empirical likelihood, the Bartlett-type corrected test statistics, with the bootstrap procedure, are proposed to achieve a higher-order accurate testing inference for the nonparametric setup as well as the parametric setup.
Access this chapter
Tax calculation will be finalised at checkout
Purchases are for personal use only
Notes
- 1.
The idea of using “estimating equations” has a long history in the literature, at least since Karl Pearson’s introduction of the so-called method of moments. It seems that a motivation behind Qin and Lawless [62] is Godambe’s optimal estimating equations theory. In the econometric literature, Hansen’s [32] generalized method of moments is a basic inferential technique. We do not mention these two big methodologies anymore, to save space.
- 2.
The finding was first announced by the author at The Mathematical Society of Japan (Spring Meeting 2012).
- 3.
The bootstrap method was introduced by Efron [25], inspired by an earlier work on the jackknife. The methodology of the bootstrap is a computer-intensive technique that provides a basis for every field from modern statistical science. It is well recognized that its theoretical validity, as well as a deep understanding of the bootstrap procedure, is closely related to higher-order asymptotic statistical theory (e.g., Hall [30]); the details are omitted here.
- 4.
A hybrid method of “the bootstrap-based Bartlett-type adjustment” from the ordinary parametric likelihood/GEL testing inference was reported by the author at Nara-Symposium (September 2014) that Professor Taniguchi organized, and also at the Japanese Joint Statistical Meeting (September 2014).
- 5.
The following properties hold:
-
If \(Y^{(N)}=o_F^{(N)}(q_1,q_2)\), then, (i) \(Y^{(N)}=o_F^{(N)}(q_1',q_2')\) for any \(q_1 \ge q_1'\)(\(\ge 0\)) and \(q_2' \ge q_2\); (ii) \(N^{-\tau } Y^{(N)}=o_F^{(N)}(q_1,q_2')\) for any \(\tau >0\) and \(q_2' \ge 0\).
-
If \(Y_\textrm{I}^{(N)}=o_F^{(N)}(q,q_2)\) and \(Y_\textrm{II}^{(N)}=o_F^{(N)}(q,q_2')\), then,
$$Y_\textrm{I}^{(N)}+Y_\textrm{II}^{(N)}=o_F^{(N)}(q,\max (q_2,q_2')) \quad \text{ and } \quad Y_\textrm{I}^{(N)} Y_\textrm{II}^{(N)}=o_F^{(N)}(q,q_2+q_2')\,. $$
-
- 6.
- 7.
The Wald-type is a GEL counterpart of Wald [75] from the parametric likelihood.
- 8.
- 9.
The gradient-type is a GEL counterpart of [73] from the parametric likelihood.
- 10.
Newey and Smith [56] established that
$$N^{1/2}(\widehat{\boldsymbol{{\eta }}}^{(N)\rho }-\boldsymbol{{\eta }}_0) {\mathop {\longrightarrow }\limits ^{d}}\textrm{N}\left( \textbf{0}_{p+M}\,,\, \left( \begin{array}{@{}c@{~}c@{}} ( \boldsymbol{{\nu }}_{\theta \lambda } \boldsymbol{{\nu }}_{\lambda ,\lambda }^{-1} \boldsymbol{{\nu }}_{\lambda \theta } )^{-1} &{} \textbf{O}_{p,M} \\ \textbf{O}_{M,p} &{} \textbf{P} \\ \end{array} \right) \right) \,, $$where \(\boldsymbol{{\nu }}_{\lambda ,\lambda }=-\boldsymbol{{\nu }}_{\lambda \lambda }\) and
$$\textbf{P} =\boldsymbol{{\nu }}_{\lambda ,\lambda }^{-1} -\boldsymbol{{\nu }}_{\lambda ,\lambda }^{-1} \boldsymbol{{\nu }}_{\lambda \theta } ( \boldsymbol{{\nu }}_{\theta \lambda } \boldsymbol{{\nu }}_{\lambda ,\lambda }^{-1} \boldsymbol{{\nu }}_{\lambda \theta } )^{-1} \boldsymbol{{\nu }}_{\theta \lambda } \boldsymbol{{\nu }}_{\lambda ,\lambda }^{-1}\,. $$We can see that
which yields \(N^{1/2}(\widehat{\boldsymbol{{\theta }}}_{(1)}^{(N)\rho }-\boldsymbol{{\theta }}_{(1)0}) {\mathop {\longrightarrow }\limits ^{d}}\textrm{N}(\textbf{0}_{p_1},\boldsymbol{{\nu }}_{(11 \cdot 2)}^{-1}) \), since
$$(\textbf{I}_{p_1}~ \textbf{O}_{p_1,p_2}) ( \boldsymbol{{\nu }}_{\theta \lambda } \boldsymbol{{\nu }}_{\lambda \lambda }^{-1} \boldsymbol{{\nu }}_{\lambda \theta } )^{-1} \left( \begin{array}{@{}c@{}} \textbf{I}_{p_1} \\ \textbf{O}_{p_2,p_1} \\ \end{array} \right) =-( \boldsymbol{{\nu }}_{\theta _{(1)}\lambda } \textbf{M}_{\lambda \lambda } \boldsymbol{{\nu }}_{\lambda \theta _{(1)}} )^{-1} =\boldsymbol{{\nu }}_{(11 \cdot 2)}^{-1}\,. $$It follows that \(\textrm{W}^{(N)\rho } {\mathop {\longrightarrow }\limits ^{d}}\chi ^2_{p_1}\) and \(\textrm{W}_\dagger ^{(N)\rho } {\mathop {\longrightarrow }\limits ^{d}}\chi ^2_{p_1}\).
- 11.
- 12.
Some authors may use the terminology of the Bartlett correction in the sense that the expectation of the test statistic is closer to that of the original (uncorrected) test statistic.
- 13.
- 14.
There were, at least for me, confusing expressions in Cordeiro and Ferrari [20, (1) and (2)]; indeed, using the relation \(2g_{\nu +2}(x)=G_\nu (x)-G_{\nu +2}(x)\), one can rearrange Harris’s [34, (3.2)] asymptotic expansion for the distribution of the Rao test statistic \(S_\textrm{R}\), as follows:
$$\begin{aligned}{} & {} \Pr (S_\textrm{R} \le x) \\= & {} G_m(x) +\frac{1}{24n}\,[ A_3 G_{m+6}(x) +(A_2-3A_3) G_{m+4}(x) +(A_1-2A_2+3A_3) G_{m+2}(x) \\{} & {} \qquad \qquad \qquad +(-A_1+A_2-A_3) G_m(x) ] +o(n^{-1}) \\= & {} G_m(x) -\frac{1}{12n} \Bigl [ \frac{A_1-A_2+A_3}{m} +\frac{(A_2-2A_3)x}{m(m+2)} +\frac{A_3\,x^2}{m(m+2)(m+4)} \Bigr ] x g_m(x) +o(n^{-1})\,. \end{aligned}$$ - 15.
- 16.
Similar conclusion holds for the modern GEL framework; indeed, Bravo [10] explicitly derived \(N^{-1/2}\)-asymptotic expansion for the distribution (under a sequence of contiguous alternatives) of a class of ECR test statistics for testing a simple full vector parameter hypothesis in the just-identified moment restrictions. To the best of our knowledge, the literature is, however, little.
- 17.
Note that
$$1-P_F^{(N)}(\mathcal{X}_0^{(N)}) \le P_F^{(N)} \Bigl [ \max _{i \in \{ 1,\ldots ,N \}} \sup _{\theta \in \Theta } ||\textbf{g}(\textbf{X}_i,\boldsymbol{{\theta }})|| > \frac{1}{2}\,N^{1/2-\xi _0} \Bigr ] =o(N^{-1}) $$(see Lemma 10.2).
- 18.
There exists an integer \(N_{0,\rho }\) such that \(\pm N^{-\xi _0} \log N \in \mathcal{N}_\rho \)(\(\subset \mathcal{V}_\rho \)) for all \(N \ge N_{0,\rho }\).
- 19.
Apply the fact (10.1) to get
$$\begin{aligned} \rho _{M+1}^\uparrow (\boldsymbol{{\nu }}+\boldsymbol{{\Psi }}^{(N)\rho }(\boldsymbol{{\eta }}))\ge & {} \rho _{M+1}^\uparrow (\boldsymbol{{\nu }})-||\boldsymbol{{\Psi }}^{(N)\rho }(\boldsymbol{{\eta }})|| \ge \frac{1}{ ||\boldsymbol{{\nu }}^{-1}|| }-||\boldsymbol{{\Psi }}^{(N)\rho }(\boldsymbol{{\eta }})|| >\frac{1}{ 2||\boldsymbol{{\nu }}^{-1}|| }\,, \\ \rho _M^\uparrow (\boldsymbol{{\nu }}+\boldsymbol{{\Psi }}^{(N)\rho }(\boldsymbol{{\eta }}))\le & {} \rho _M^\uparrow (\boldsymbol{{\nu }})+||\boldsymbol{{\Psi }}^{(N)\rho }(\boldsymbol{{\eta }})|| \le -\frac{1}{ ||\boldsymbol{{\nu }}^{-1}|| }+||\boldsymbol{{\Psi }}^{(N)\rho }(\boldsymbol{{\eta }})|| <-\frac{1}{ 2||\boldsymbol{{\nu }}^{-1}|| }\,, \end{aligned}$$since \(\min \{ \rho _{M+1}^\uparrow (\boldsymbol{{\nu }}),-\rho _M^\uparrow (\boldsymbol{{\nu }}) \} =1/||\boldsymbol{{\nu }}^{-1}|| \) (see (10.6)).
- 20.
Some straightforward but tedious calculations show that
$$\begin{aligned} \kappa _{a_1}^{\rho _3,\tau _1}= & {} -\nu ^\mathcal{G}_{a_1r,r'} \nu _{(22)}^{rr'} +\frac{1}{2}\,{\nu ^\rho }^\mathcal{G}_{a_1r_1r_2} \nu _{(22)}^{r_1r_1'} \nu _{r_1',r_2'} \nu _{(22)}^{r_2r_2'} \\{} & {} +\frac{1}{2} \Bigl ( \tau _1 {\nu ^\rho }^\mathcal{G\,\,G\,\,G}_{a_1b\,b'} -\nu ^\mathcal{G\,\,G\,\,G}_{a_1b,b'} -\sum _{\beta '=1}^M \mathcal{G}_{[\beta ']b'} {\nu ^\rho }^\mathcal{G\,\,G}_{a_1b[\beta ']} \Bigr ) \nu _{(11 \cdot 2)}^{bb'}\,, \end{aligned}$$and that \(\kappa _{a_1,a_2,a_3}^{\rho _3,\tau _1}\) and \(\kappa _{a_1,a_2,a_3,a_4}^{-2,\rho _4,1/3,\tau _2,\tau _3,\tau _4}\) are explicitly given in the proof of Theorem 10.1; the lengthy general expression for \(\kappa _{a_1,a_2,a_3,a_4}^{\rho _3,\rho _4,\tau _1,\tau _2,\tau _3,\tau _4}\) is, however, omitted here. Although, in principle, we can write down \(\kappa _{a_1,a_2}^{\rho _3,\rho _4,\tau _1,\tau _2,\tau _3,\tau _4}\), we have not rearranged it, due to the rather lengthy algebra (the task would be no practical importance, in many cases).
References
Anderson, T. W. (2003). An Introduction to Multivariate Statistical Analysis, 3rd ed. New York: Wiley.
Baggerly, K.A. (1998). Empirical likelihood as a goodness-of-fit measure. Biometrika 85 535–547.
Barndorff-Nielsen, O.E. and Cox, D.R. (1994). Inference and Asymptotics. London: Chapman and Hall.
Bartlett, M.S. (1937). Properties of sufficiency and statistical tests. Proc.Roy.Soc.Lond. A 160 268–282.
Bera, A.K. and Bilias, Y. (2001). Rao’s score, Neyman’s \(C(\alpha )\) and Silvey’s LM tests: an essay on historical developments and some new results. J.Statist.Plann.Inference 97 9–44.
Bhattacharya, R.N. and Ghosh, J.K. (1978). On the validity of the formal Edgeworth expansion. Ann.Statist. 6 434–451. Correction: (1980). 8 1399.
Bhattacharya, R.N. and Rao, R.R. (1976). Normal Approximation and Asymptotic Expansions. New York: Wiley.
Boos, D.D. (1992). On generalized score tests. Amer.Statist. 46 327–333.
Bravo, F. (2002). Testing linear restrictions in linear models with empirical likelihood. Econom.J. 5 104–130.
Bravo, F. (2003). Second-order power comparisons for a class of nonparametric likelihood-based tests. Biometrika 90 881–890.
Bravo, F. (2004). Empirical likelihood based inference with applications to some econometric models. Econometric Theory 20 231–264.
Bravo, F. (2006). Bartlett-type adjustments for empirical discrepancy test statistics. J.Statist.Plann.Inference 136 537–554.
Camponovo, L. and Otsu, T. (2014). On Bartlett correctability of empirical likelihood in generalized power divergence family. Statist.Probab.Lett. 86 38–43.
Chandra, T.K. and Mukerjee, R. (1991). Bartlett-type modification for Rao’s efficient score statistic. J.Multivariate Anal. 36 103–112.
Chen, S.X. (1993). On the accuracy of empirical likelihood confidence regions for linear regression model. Ann.Inst.Statist.Math. 45 621–637.
Chen, S.X. (1994). Empirical likelihood confidence intervals for linear regression coefficients. J.Multivariate Anal. 49 24–40.
Chen, S.X. and Cui, H. (2006). On Bartlett correction of empirical likelihood in the presence of nuisance parameters. Biometrika 93 215–220.
Chen, S.X. and Cui, H. (2007). On the second-order properties of empirical likelihood with moment restrictions. J.Econometrics 141 492–516.
Corcoran, S.A. (1998). Bartlett adjustment of empirical discrepancy statistics. Biometrika 85 967–972.
Cordeiro, G.M. and Ferrari, S.L.P. (1991). A modified score test statistic having chi-squared distribution to order \(n^{-1}\). Biometrika 78 573–582.
Cox, D.R. (1988). Some aspects of conditional and asymptotic inference: a review. Sankhyā A 50 314–337.
Cressie, N. and Read, T. (1984). Multinomial goodness-of-fit tests. J.R.Stat.Soc.Ser. B 46 440–464.
Cribari-Neto, F. and Cordeiro, G.M. (1996). On Bartlett and Bartlett-type corrections. Econometric Rev. 15 339–367.
DiCiccio, T., Hall, P. and Romano, J. (1991). Empirical likelihood is Bartlett-correctable. Ann.Statist. 19 1053–1061.
Efron, B. (1979). Bootstrap methods: another look at the jackknife. Ann.Statist. 7 1–26.
Fisher, R.A. (1922). On the mathematical foundations of theoretical statistics. Philos.Trans.R.Soc.Lond.Ser.A Math.Phys.Eng.Sci. 222 309–368.
Fujikoshi, Y. (1988). Comparison of powers of a class of tests for multivariate linear hypothesis and independence. J.Multivariate Anal. 26 48–58.
Fujikoshi, Y. (1997). A method for improving the large-sample chi-squared approximations to some multivariate test statistics. Amer.J.Math.Management Sci. 17 15–29.
Fujisawa, H. (1997). Improvement on chi-squared approximation by monotone transformation. J.Multivariate Anal. 60 84–89.
Hall, P. (1992). The Bootstrap and Edgeworth Expansion. New York: Springer.
Hall, P. and La Scala, B. (1990). Methodology and algorithms of empirical likelihood. Internat.Statist.Rev. 58 109–127.
Hansen, L.P. (1982). Large sample properties of generalized method of moments estimators. Econometrica 50 1029–1054.
Hansen, L.P., Heaton, J. and Yaron, A. (1996). Finite-sample properties of some alternative GMM estimators. J.Bus.Econom.Statist. 14 262–280.
Harris, P. (1985). An asymptotic expansion for the null distribution of the efficient score statistic. Biometrika 72 653–659.
Harris, P. and Peers, H.W. (1980). The local power of the efficient scores test statistic. Biometrika 67 525–529.
Hayakawa, T. (1975). The likelihood ratio criterion for a composite hypothesis under a local alternative. Biometrika 62 451–460.
Hayakawa, T. (1977). The likelihood ratio criterion and the asymptotic expansion of its distribution. Ann.Inst.Statist.Math. 29 359–378. Correction: (1987). 39 681.
Horn, R.A. and Johnson, C.R. (1985). Matrix Analysis. Cambridge: Cambridge University Press.
Iwashita, T. (1997). Adjustment of Bartlett type to Hotelling’s \(T^2\)-statistic under elliptical distributions. Amer.J.Math.Management Sci. 17 91–96.
Jensen, J.L. (1993). A historical sketch and some new results on the improved log likelihood ratio statistic. Scand.J.Statist. 20 1–15.
Jing, B.Y. and Wood, A.T.A. (1996). Exponential empirical likelihood is not Bartlett correctable. Ann.Statist. 24 365–369.
Kakizawa, Y. (1996). Higher order monotone Bartlett-type adjustment for some multivariate test statistics. Biometrika 83 923–927.
Kakizawa, Y. (1997). Higher-order Bartlett-type adjustment. J.Stat.Plan.Inference 65 269–280.
Kakizawa, Y. (2009). Third-order power comparisons for a class of tests for multivariate linear hypothesis under general distributions. J.Multivariate Anal. 100 473–496.
Kakizawa, Y. (2012a). Generalized Cordeiro–Ferrari Bartlett-type adjustment. Statist.Probab.Lett. 82 2008–2016.
Kakizawa, Y. (2012b). Improved chi-squared tests for a composite hypothesis. J.Multivariate Anal. 107 141–161.
Kakizawa, Y. (2013). Third-order local power properties of tests for a composite hypothesis. J.Multivariate Anal. 114 303–317.
Kakizawa, Y. (2015). Third-order local power properties of tests for a composite hypothesis, II. J.Multivariate Anal. 140 99–112.
Kakizawa, Y. (2017). Third-order average local powers of Bartlett-type adjusted tests: Ordinary versus adjusted profile likelihood. J.Multivariate Anal. 153 98–120.
Kitamura, Y. and Stutzer, M. (1997). An information-theoretic alternative to generalized method of moments estimation. Econometrica 65 861–874.
Kojima, M. and Kubokawa, T. (2013). Bartlett-type adjustments for hypothesis testing in linear models with general error covariance matrices. J.Multivariate Anal. 122 162–174.
Kolaczyk, E.D. (1994). Empirical likelihood for generalized linear models. Statist.Sinica 4 199–218.
Lawley, D.N. (1956). A general method for approximating to the distribution of likelihood ratio criteria. Biometrika 43 295–303.
Magdalinos, M.A. (1992). Stochastic expansions and asymptotic approximations. Econometric Theory 8 343–367.
Mukerjee, R. (1993). Rao’s score test: Recent asymptotic results. In: Handbook of Statistics, Vol.11. Maddala, G.S., Rao, C.R. and Vinod, H.D. Eds. Amsterdam: North-Holland, pp.363–379.
Newey, W.K. and Smith, R.J. (2004). Higher order properties of GMM and generalized empirical likelihood estimators. Econometrica 72 219–255.
Owen, A.B. (1988). Empirical likelihood ratio confidence intervals for a single functional. Biometrika 75 237–249.
Owen, A.B. (1990). Empirical likelihood ratio confidence regions. Ann.Statist. 18 90–120.
Owen, A.B. (1991). Empirical likelihood for linear models. Ann.Statist. 19 1725–1747.
Owen, A.B. (2001). Empirical Likelihood. London: Chapman and Hall.
Peers, H.W. (1971). Likelihood ratio and associated test criteria. Biometrika 58 577–587.
Qin, J. and Lawless, J. (1994). Empirical likelihood and general estimating equations. Ann.Statist. 22 300–325.
Qin, J. and Lawless, J. (1995). Estimating equations, empirical likelihood and constraints on parameters. Canad.J.Statist. 23 145–159.
Ragusa, G. (2011). Minimum divergence, generalized empirical likelihoods, and higher order expansions. Econometric Rev. 30 406–456.
Rao, C.R. (1948). Large sample tests of statistical hypotheses concerning several parameters with applications to problems of estimation. Math.Proc.Cambridge Philos.Soc. 44 50–57.
Rao, C.R. and Mukerjee, R. (1997). Comparison of LR, score, and Wald tests in a non-iid setting. J.Multivariate Anal. 60 99–110.
Rocke, D.M. (1989). Bootstrap Bartlett adjustment in seemingly unrelated regression. J.Amer.Statist.Assoc. 84 598–601.
Rothenberg, T.J. (1977). Edgeworth expansions for multivariate test statistics. IP–255, Center for Research in Management Science, University of California, Berkeley.
Silvey, S.D. (1959). The Lagrange multiplier test. Ann.Math.Statist. 30 389–407.
Siotani, M., Hayakawa, T. and Fujikoshi, Y. (1985). Modern Multivariate Analysis: A Graduate Course and Handbook. Ohio: American Sciences Press.
Taniguchi, M. (1991a). Higher Order Asymptotic Theory for Time Series Analysis. Berlin: Springer-Verlag.
Taniguchi, M. (1991b). Third-order asymptotic properties of a class of test statistics under a local alternative. J.Multivariate Anal. 37 223–238.
Terrell, G.R. (2002). The gradient statistic. Comput.Sci.Statist. 34 206–215.
Thomas, D.R. and Grunkemeier, G.L. (1975). Confidence interval estimation of survival probabilities for censored data. J.Amer.Statist.Assoc. 70 865–871.
Wald, A. (1943). Tests of statistical hypothesis concerning several parameters when the number of observations is large. Trans.Amer.Math.Soc. 54 426–482.
Wilks, S.S. (1938). The large-sample distribution of the likelihood ratio for testing composite hypotheses. Ann.Math.Statist. 9 60–62.
Zhang, B. (1996). On the accuracy of empirical likelihood confidence intervals for \(M\)-functionals. J.Nonparametr.Stat. 6 311–321.
Author information
Authors and Affiliations
Corresponding author
Editor information
Editors and Affiliations
Rights and permissions
Copyright information
© 2023 The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
About this chapter
Cite this chapter
Kakizawa, Y. (2023). Asymptotic Expansions for Several GEL-Based Test Statistics and Hybrid Bartlett-Type Correction with Bootstrap. In: Liu, Y., Hirukawa, J., Kakizawa, Y. (eds) Research Papers in Statistical Inference for Time Series and Related Models. Springer, Singapore. https://doi.org/10.1007/978-981-99-0803-5_10
Download citation
DOI: https://doi.org/10.1007/978-981-99-0803-5_10
Published:
Publisher Name: Springer, Singapore
Print ISBN: 978-981-99-0802-8
Online ISBN: 978-981-99-0803-5
eBook Packages: Mathematics and StatisticsMathematics and Statistics (R0)