Abstract

We present some shortcomings of the standard t-test when it is applied to multiple comparisons. We propose an algorithm (Dubey/Armitage-Parmar) that represents a better compromise between statistical errors of type I and type II.

© 2008 Optical Society of America

PDF Article