Using recently published examples, two main theoretical advantages of a rec
ently proposed modification of Malinowski's F-test are illustrated, i.e., (
1) sharper significance levels are obtained, and (2) the new test statistic
better approaches a value of one under the null-hypothesis. A serious prac
tical drawback of the modification, namely that it is more complicated than
the original test, is alleviated by supplying carefully documented Matlab
code. (C) 1999 Elsevier Science Ltd. All rights reserved.