Geodesic
Calculation of power measures of Grabbs' criterion for checking for one outlier
Sibirskij žurnal industrialʹnoj matematiki, Tome 13 (2010) no. 4, pp. 141-154 
L. K. Shiryaeva. Calculation of power measures of Grabbs' criterion for checking for one outlier. Sibirskij žurnal industrialʹnoj matematiki, Tome 13 (2010) no. 4, pp. 141-154. http://geodesic.mathdoc.fr/item/SJIM_2010_13_4_a12/
@article{SJIM_2010_13_4_a12,
     author = {L. K. Shiryaeva},
     title = {Calculation of power measures of {Grabbs'} criterion for checking for one outlier},
     journal = {Sibirskij \v{z}urnal industrialʹnoj matematiki},
     pages = {141--154},
     year = {2010},
     volume = {13},
     number = {4},
     language = {ru},
     url = {http://geodesic.mathdoc.fr/item/SJIM_2010_13_4_a12/}
}
TY  - JOUR
AU  - L. K. Shiryaeva
TI  - Calculation of power measures of Grabbs' criterion for checking for one outlier
JO  - Sibirskij žurnal industrialʹnoj matematiki
PY  - 2010
SP  - 141
EP  - 154
VL  - 13
IS  - 4
UR  - http://geodesic.mathdoc.fr/item/SJIM_2010_13_4_a12/
LA  - ru
ID  - SJIM_2010_13_4_a12
ER  - 
%0 Journal Article
%A L. K. Shiryaeva
%T Calculation of power measures of Grabbs' criterion for checking for one outlier
%J Sibirskij žurnal industrialʹnoj matematiki
%D 2010
%P 141-154
%V 13
%N 4
%U http://geodesic.mathdoc.fr/item/SJIM_2010_13_4_a12/
%G ru
%F SJIM_2010_13_4_a12

Voir la notice de l'article provenant de la source Math-Net.Ru

We obtain recurrences for calculating five measures of the David power of Grabbs' criterion in the case that a normally distributed sample contains an outlier. We show that the power measures are functions of the parameters of the outlier, the sample volume, and the critical values of the Grabbs statistics. We proved that all power measures, except for the fourth, are nonincreasing functions of the critical values of the Grabbs statistics, while the fourth power measure always has a local maximum. Using our formulas we ran model simulations of the power measures in the case of a normally distributed sample of 20 observations including one outlier. The results of the simulations turn out close to those expected theoretically.

[1] Deivid G., Poryadkovye statistiki, Nauka, M., 1979 | MR

[2] Grubbs F., “Sample criteria for testing outlying observations”, Ann. Math. Statist., 21:1 (1950), 27–58 | DOI | MR | Zbl

[3] David H. A., Paulson A. S., “The performance of several tests for outliers”, Biometrica, 52 (1965), 429–436 | MR | Zbl

[4] Dixon W. J., “Analysis of extreme values”, Ann. Math. Statist., 21:4 (1950), 488–506 | DOI | MR | Zbl

[5] Lemeshko B. Yu., Lemeshko S. B., “Rasshirenie oblasti primeneniya kriteriev tipa Grabbsa, ispolzuemykh pri otbrakovke anomalnykh nablyudenii”, Izmeritelnaya tekhnika, 2005, no. 6, 13–19

[6] Livesey J. H., “Curtosis provides a good omnibus test for outliers in small samples”, Clinical Biochem., 40 (2007), 1032–1036 | DOI

[7] Aivazyan S. A., Mkhitaryan V. S., Teoriya veroyatnostei i prikladnaya statistika, v. 1, YuNITI-DANA, M., 2001

[8] Zhang J., Keming Y., “The null distribution of the likelihood-ratio test for one or two outliers in a normal sample”, Sociedad de Estatistica e Investigacion Operativa Test, 15:1 (2006), 141–150 | MR | Zbl

[9] Barnett V., Lewis T., Outliers in statistical data, John Wiley and Sons, Chichester, 1984 | MR | Zbl