7: Further Topics on Logistic Regression

Overview Section

In the previous lesson, we dealt with basic topics of logistic regression. Recall that logistic regression is a special type of regression where the probability of "success" is modeled through a set of predictors. The predictors may be categorical, nominal or ordinal, or continuous. Logit or log odds of success is assumed to be a linear function of the predictors and such a model is fitted. This lesson deals with continuous covariates and model diagnostics for logistic regression.

Upon completion of this lesson, you should be able to:

  Objective 7.1

Explain the role that logistic regression model residuals play in assessing the model fit and use them to measure evidence against the model.

  Objective 7.2

Assess the relative importance of multiple predictors when fitting a logistic regression model

  Objective 7.3

Explain what overdispersion means in the logistic regression model and account for it when making conclusions.

  Objective 7.4

Interpret sensitivity and specificity in the context of logistic regression and use them to assess the predictive power of the model.

 Lesson 7 Code Files

Data File: donner.txt