Advances and Applications in Statistics
Volume 39, Issue 2, Pages 81 - 94
(April 2014)
|
|
IMPROVING ACCURACY OF TEST OF INDEPENDENCE
Jan Vrbik
|
Abstract: In this article, we propose to improve the accuracy of the usual test of independence of two nominal-scale attributes by finding a -proportional correction to the chi-squared distribution of the corresponding test statistics. This correction is based on Edgeworth expansion of the multivariate distribution of observed frequencies, adapted to approximate the distribution of their standardized sum of squares. |
Keywords and phrases: independence test, Edgeworth series, asymptotic distribution. |
|
Number of Downloads: 373 | Number of Views: 1199 |
|