4.7 Chapter 4 Major Takeaways

Logistic regression is used to predict binary categorical outcomes based on explanatory variables.

• The benefit of using logistic regression in this setting over linear regression is that we constrain the predicted probabilities to be between 0 and 1.
• When interpreting logistic regression models, you need to exponentiate the slope coefficients to get meaning values to interpret. The exponentiate slopes are ratios of odds of the outcome, compare two groups defined by the explanatory variable (quantitative X: groups are 1 unit apart in X; categorical X: indicator group compared to reference level).
• To evaluate the model, we consider the predicted probabilities within groups based on the true outcome. We hope that the predicted probabilities given from the model are well separated between the groups based on true outcome values.
• Additionally, to evaluate the model, we convert those probabilities to predicted outcomes by setting a threshold from which the accuracy, sensitivity, specificity, false positive rate, and false negative rate can be calculated.