ACR

Upper percentiles squared Mahalanobis distance critical value for test of single multivariate normal

Description

From the method given by Wilks (1963) and approaching to a F distribution function by the Yang and Lee (1987) formulation, we provide an m-file to get the critical value of the maximun squared Mahalanobis distance to detect outliers from a normal multivariate sample.

--The function's name is giving as a gratefull to Dr. Alvin C. Rencher for his unvaluable contribution to the teaching of multivariate statistics with his text 'Methods of Multivariate Analysis'.--
     
Inputs:
    p - number of independent variables.
    n - sample size.
alpha - significance level (default = 0.05).

Output:
      x - critical value of the maximun squared Mahalanobis distance.

We can generate all the critical values of the maximun squared Mahalanobis distance presented on the Table XXXII of by Barnett and Lewis (1978) and Table A.6 of Rencher (2002). Also with any given significance level (alpha).

• MATLAB release: MATLAB 7 (R14)

[Andreas Procopiou]