STATISTICAL CONSIDERATIONS BASED ON AN INSTRUMENT FOR MEASURING ADVISER PERCEPTION

An instrument is described for measuring the degree to which an adviser perceives' his advisees' attitudes. Its application requires each of n advisees to respond to a form consisting of c items. For each item, the adviser is instructed to respond as he thinks the particular advisee under consideration did. If X ij = j th advisee's response to the i th item, and if Y ij = adviser's estimate of X ij , then θ ij = | X ij – Y ij | is a reasonable measure of response discrepancy. Since the distribution of θ ij for a given patient response X ij = x depends on x , an ‘equitable’ comparison between two advisers must take into account differences in the distributions of advisee responses. A ‘standardized’ perception score is developed which reflects the fractional gain in perception from its worst possible value, conditional on the observed set of advisee responses. A method for testing the null hypothesis that two advisers are equally perceptive is given and a numerical example is included which illustrates the necessity for standardizing.