For well over a decade statistical program packages have been providing standard routines for fitting probit and logit models, and these methods have become commonplace tools of applied statistical research. The logit model is the more versatile of the two: its simple and elegant analytical properties permit its use in widely different contexts and for a variety of purposes. This monograph treats logistic regression as the core of a number of such variations and generalizations. Its purpose is to present several widely used models that are based on the logit transformation and to answer the questions that arise in the practice of empirical research. It explains the theoretical background of these models, their estimation and some further statistical analysis, the interpretation of the results, and their practical applications, all at an elementary level. I assume that readers are familiar with ordinary linear regression and with the estimation theory and matrix algebra that go with it.

Parts of the book are taken from The Logit Model: an Introduction for Economists, published in 1989 by Edward Arnold. One of the things I have learned since then is that several varieties of logit analysis have been developed independently, in almost perfect isolation, in various branches of biology, medicine, economics, and still other disciplines. In each field practitioners use distinct approaches, interpretations and terminologies of their own. I have tried to overcome these differences and to do justice to the main developments in other fields, but this will not conceal that my own background and inspiration are in econometrics.