A Consistent Nonparametric Test for Causality in Quantile



σ=2E{σε(z,)f-.(z,)}Kuu)du and σ'ε<z,)=E(4z<)(1-θ).

(ii) under the null hypothesis (3), Cf^ 2θ2(1 - θ)2------------ Ktis is a consistent

T(T- 1)hm st

estimator of σ02= 2E {σε4(zt) fz(zt)} K2(u)du.

(iii) under the alternative hypothesis (4), JT p E{[Fy,(Qθ(xt) zt)-θ]2 fz(zt)} 0.

m/22

(iv) under the local alternatives (17), Th JTN(μ,σ0) in distribution, where

μ = E [. ⅛, { Qθ (zt. )iɪt} 12(,ε ) f, (,t) )].

Theorem 1 generalizes the results of Zheng (1998) of independent data to the weakly
dependent data case. A detailed proof of Theorem is given in the Appendix. The main
difficulty in deriving the asymptotic distribution of the statistic defined in equation (12) is that
a nonparametric quantile estimator is included in the indicator function which is not
differentiable with respect to the quantile argument and thus prevents taking a Taylor
expansion around the true conditional quantile
Qθ(xt). To circumvent the problem, Zheng
(1989) appealed to the work of Sherman (1994) on uniform convergence of U-statistics
indexed by parameters. In this paper, we bound the test statistic by the statistics in which the
nonparametric quantile estimator in the indicator function is replaced with sums of the true
conditional quantile and upper and lower bounds consistent with uniform convergence rate of
the nonparametric quantile estimator,
1(ytQθ(xt)-CT) and 1(ytQθ(xt)+CT).

An important further step is to show that the differences of the ideal test statistic JT given
in equation (8) and the statistics having the indicator functions obtained from the first step
stated above is asymptotically negligible. We may directly show that the second moments of
the differences are asymptotically negligible by using the result of Yoshihara (1976) on the
bound of moments of U-statistics for absolutely regular processes. However, it is tedious to
get bounds on the second moments with dependent data. In the proof we instead use the fact
that differences are second-order degenerate U-statistics. Thus by using the result on the
asymptotic normal distribution of the second-order degenerate U-statistic of Fan and Li



More intriguing information

1. Gender and aquaculture: sharing the benefits equitably
2. Endogenous Heterogeneity in Strategic Models: Symmetry-breaking via Strategic Substitutes and Nonconcavities
3. TINKERING WITH VALUATION ESTIMATES: IS THERE A FUTURE FOR WILLINGNESS TO ACCEPT MEASURES?
4. The name is absent
5. The name is absent
6. Implementation of Rule Based Algorithm for Sandhi-Vicheda Of Compound Hindi Words
7. Work Rich, Time Poor? Time-Use of Women and Men in Ireland
8. Der Einfluß der Direktdemokratie auf die Sozialpolitik
9. The name is absent
10. ‘Goodwill is not enough’
11. Testing Gribat´s Law Across Regions. Evidence from Spain.
12. Evaluating Consumer Usage of Nutritional Labeling: The Influence of Socio-Economic Characteristics
13. Regional science policy and the growth of knowledge megacentres in bioscience clusters
14. The name is absent
15. Keynesian Dynamics and the Wage-Price Spiral:Estimating a Baseline Disequilibrium Approach
16. Migrating Football Players, Transfer Fees and Migration Controls
17. Correlates of Alcoholic Blackout Experience
18. The name is absent
19. Regional dynamics in mountain areas and the need for integrated policies
20. The name is absent