BrunnerMunzelStat_js

This function estimates value of statistic used in "Brunner Munzel test" with given two sample arrays.
https://github.com/YujiSODE/BrunnerMunzelStat_js

Copyright (c) 2017 Yuji SODE <yuji.sode@gmail.com>
This software is released under the MIT License.
See LICENSE or http://opensource.org/licenses/mit-license.php

---

According to Brunner and Munzel (2000), a value of estimated statistic in this method follows the standard normal distribution asymptotically; this distribution can be approximated to t-distribution when n₁ and n₂ are not less than 10,
while whole sample size N =n₁+n₂.

Script

• BrunnerMunzelStat_js.js
  Function: BMStat(X,Y); X and Y are numerical arrays.

This function rettuns object with three values: statistic, estP and df.

• statistic: value of statistic
• estP: estimated value P = P(X<Y)+P(X=Y)/2
• df: degrees of freedom for approximation by t-distribution

Example data cited from Brunner and Munzel (2000)

var v1=[1,2,1,1,1,1,1,1,1,1,2,4,1,1],v2=[3,3,4,3,1,2,3,1,1,5,4];
BMStat(v1,v2);
Result: { statistic: 3.1374674823029514, estP: 0.7889610389610391, df: 17.682841979481548 }

References and Library

References cited

• Brunner, E. and Munzel, U. 2000. The Nonparametric Behrens-Fisher Problem: Asymptotic Theory and a Small-Sample Approximation. Biometrical Jour. vol. 42, p. 17-25.
• R Documentation brunner.munzel.test derived on [2017-06-19]
• Brunner-Munzel test (in Japanese) derived on [2017-06-19]

Library

• stRank.js (Yuji SODE,2017): the MIT License; https://gist.github.com/YujiSODE/f913c0c2cf49353e561cd03c2a7a8a6b