| This book is motivated to a large extent by our dissatisfaction with current practices in behavioral measurement. Most of the basic data consists of dichotomous responses, and much of the rest is made of responses on short scales. Neither type of data furnishes information that can inherently be considered more than ordinal. The dominant contemporary treatment of this data is to derive from it scores on interval-scale latent variables through the application of one or another"model" that is presumed to explain the data. We feel that, in the majority of instances, the application of these models is inappropriate: The models either do not fit the data, or are even inherently self-contradictory. |