×

A nonparametric Hellinger metric test for conditional independence. (English) Zbl 1284.62285

Summary: We propose a nonparametric test of conditional independence based on the weighted Hellinger distance between the two conditional densities, \(f(y|x,z)\) and \(f(y|x)\), which is identically zero under the null. We use the functional delta method to expand the test statistic around the population value and establish asymptotic normality under \(\beta\)-mixing conditions. We show that the test is consistent and has power against alternatives at distance \(n^{-1/2}h^{-d/4}\). The cases for which not all random variables of interest are continuously valued or observable are also discussed. Monte Carlo simulation results indicate that the test behaves reasonably well in finite samples and significantly outperforms some earlier tests for a variety of data generating processes. We apply our procedure to test for Granger noncausality in exchange rates.

MSC:

62G10 Nonparametric hypothesis testing
PDFBibTeX XMLCite
Full Text: DOI

References:

[1] DOI: 10.1023/A:1004193117918 · Zbl 0959.62034
[2] DOI: 10.1214/aos/1013203462 · Zbl 1043.62032
[3] DOI: 10.1111/1468-0262.00142 · Zbl 1022.62013
[4] Newey, Econometric Theory 10 pp 233– (1994)
[5] DOI: 10.1080/07474939608800353 · Zbl 0893.62034
[6] DOI: 10.1214/aoms/1177705055 · Zbl 0139.36301
[7] DOI: 10.1016/S0167-7152(00)00091-2 · Zbl 0966.62029
[8] DOI: 10.1214/aos/1176342558 · Zbl 0275.62033
[9] DOI: 10.1111/j.1467-9892.1996.tb00294.x · Zbl 0876.62075
[10] DOI: 10.1016/0165-1765(95)00791-1 · Zbl 0875.90178
[11] DOI: 10.1016/0304-4076(75)90032-9 · Zbl 0307.62068
[12] DOI: 10.1016/S0165-1765(97)00022-0 · Zbl 0902.90020
[13] DOI: 10.1016/S0304-4076(01)00104-X · Zbl 1051.62017
[14] Andrews, Econometric Theory 11 pp 560– (1995)
[15] DOI: 10.1016/S0304-4149(96)00096-8 · Zbl 0885.62045
[16] DOI: 10.1016/S0304-4076(01)00091-4 · Zbl 1004.62042
[17] DOI: 10.1016/S0304-4076(98)00087-6 · Zbl 0929.62054
[18] DOI: 10.2307/2951556 · Zbl 0783.62100
[19] DOI: 10.1111/1468-0262.00207 · Zbl 1017.62012
[20] DOI: 10.2307/2951582 · Zbl 0761.62166
[21] DOI: 10.1111/j.1468-0262.2005.00597.x · Zbl 1152.91729
[22] DOI: 10.1016/S0304-4076(01)00043-4 · Zbl 1053.62118
[23] DOI: 10.1214/aoms/1177730150 · Zbl 0032.42001
[24] DOI: 10.2307/2329266
[25] DOI: 10.2307/1913937 · Zbl 0289.90003
[26] DOI: 10.1214/aos/1176349403 · Zbl 0795.62036
[27] Yoshihara, Proceedings of the Fourth Prague Symposium on Asymptotic Statistics pp 539– (1989)
[28] DOI: 10.1016/0047-259X(84)90044-7 · Zbl 0528.62028
[29] DOI: 10.1007/BF00532676 · Zbl 0314.60028
[30] DOI: 10.1086/260267
[31] DOI: 10.2307/2532617
[32] DOI: 10.1016/0165-1889(80)90069-X
[33] White, Cointegration, Causality, and Forecasting: A Festschrift in Honor of Clive W.J. Granger pp 326– (1999)
[34] DOI: 10.1006/jmva.2000.1950 · Zbl 1081.62529
[35] Tenreiro, Portugaliae Mathematica 54 pp 197– (1997)
[36] DOI: 10.1093/biomet/80.3.591 · Zbl 0790.62044
[37] Gasser, Journal of the Royal Statistical Society Series B 47 pp 238– (1985)
[38] DOI: 10.2307/1912602 · Zbl 0483.60023
[39] DOI: 10.1016/0167-7152(87)90072-1 · Zbl 0635.62028
[40] DOI: 10.2307/2171962 · Zbl 0856.90020
[41] DOI: 10.1214/aos/1176342996 · Zbl 0325.62030
[42] DOI: 10.1007/BF02306030 · Zbl 0569.62097
[43] DOI: 10.2307/2298005 · Zbl 0719.62055
[44] DOI: 10.1080/10485259908832783 · Zbl 0953.62094
[45] DOI: 10.2307/1912705 · Zbl 0647.62100
[46] DOI: 10.1080/10485259908832762 · Zbl 0974.62044
[47] Pitman, Some Basic Theory for Statistical Inference (1979) · Zbl 0442.62002
[48] DOI: 10.2307/2171848 · Zbl 0854.62038
[49] DOI: 10.1016/0304-4076(88)90049-8 · Zbl 0668.62013
[50] Dawid, Journal of the Royal Statistical Society, Series B 41 pp 1– (1979)
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.