A goodness-of-fit test used in logistic regression, particularly when there are continuous covariates. Units are divided into deciles based on predicted probabilities and a 2 × 10 contingency table of expected and observed frequencies for the deciles obtained. The Hosmer-Lemeshow test statistic is the Pearson chi-square statistic for this contingency table and has an approximate chi-squared distribution with 8 degrees of freedom if the model is correct. The test has limited power.