R provides powerful statistical tools for credit scoring. We emphasize here some of the most common techniques, namely probability default estimation with logit and probit regressions and ROC curve analysis. During both behavioral and application credit scoring, one can estimate or score the probability of default in the usual way that the theory of cross-sectional econometrics suggests.

Logit and probit regressions are generalized linear regression models with binary, dependent variables, where the two outcomes can be, for example, either defaulted or not. Logit regression uses logistic function; the probit model applies a cumulative distribution function of the standard normal distribution for estimating the probability of default. Coefficients of independent variables are typically estimated by the maximum likelihood method in both cases. Logit and probit regression models can be called with the glm command, which is the generalized linear model...