kirsch@psych.toronto.edu (Sofia Kirschner) (11/28/89)
Meeting of the
Psychological Measurement and Assessment
Interest Group
Friday, December 1 @ 2:30 p.m. in Room 570, SSH
Detection-Theory Measures of Response Bias
C. Douglas Creelman
Abstract
In experiments which measure sensitivity or choice, Detection Theory has
provided measures (primarily d' of Signal Detectability Theory, and alpha of
Choice Theory), which are unaffected by the decision-maker's criterion. The
converse is not true; we have no robust measure of criterion which remains
constant with changes in discriminability. Alternative measures will be
discussed and compared, and the relatively meagre experimental literature will
be surveyed. An argument will be made that no measure is "non-parametric"
and the argument is supported by a demonstration that one proposed measure
of bias which is purported to be parametric, and one which is purported to be
non-parametric are equivalent. A strong recommendation will be made in
favor of the measure c, instead of the measure beta, for characterizing bias.