ABSTRACT
In the case when~ m = 2, with = Cjf. = j = 1, 2, this criterion takes the simple form: reject the null hypothesis when the distance betu•een th~ two estimates is larger than the .mm of the two corresponding "critical po1:nt." The rejection region is then
Form > 2, we reject the null hypotlwsis when at least one of the distances iGk 01 I is larger than the sum c1 (J;, ai) + q (y k, ak). We will now extend this procedure to multidimensional parameters and consider confidence ellipsoids.
