Logistic regression

Binary logistic regression predicts something with exactly two values (for example yes or no). It works not with the probability but with the odds: probability divided by one minus probability. The model test χ² checks whether it predicts better than a model without predictors. Nagelkerke R² (0 to 1) is a pseudo R²: not a share of differences explained like the R² of linear regression, and there are no fixed cut-offs for small or large.

Example: Phone in bed from concentration, age and coffee: χ²(3) = 11.24, p = .010, Nagelkerke R² = .12. 67% of the people were classified correctly; without a prediction it would be 64%.

Caution: Rule of thumb: at least 10 cases per predictor in the smaller group. A high hit rate is worth little if one group is much larger anyway - that is why the comparison is shown.