Practical Introduction to Logistic Regression using Excel

The Logistic regression is a nonlinear regression model that enables predictions of the probability of observing an outcome, including a binary result (e.g. YES/NO, ON/OFF). Because of this, it is also used as a classification method.

In this course, participants learn the steps required to compute the parameters of a logistic model. Students will use Excel to implement the steps required to compute the parameters of a logistic model using one or more predictor variables.

Guided exercises support the students to ensure practical understanding and immediate applicability of the concepts.

Introduction to regression and the binary prediction problem (4 hours)

• The limitations of a linear probability model,
• The differences between linear and logistic regression
• The logarithm: properties and use in logistic regression

Logistic function, logit transformation

• The odds and logarithm of the odds.
• Computing the model parameters for a single predictor: the log-likelihood function

Interpreting the model parameters 

• The odds ratio and the model slope parameter
• Quantifying model performance:  the confusion matrix and fraction of correct responses
• Model significance: The Chi-squared test and comparing against an intercept-only model

• State when logistic regression models are needed, and the weakness of a linear probability model.
• Explain the logit transformation and its purpose
• Apply SOLVER to estimate the parameters of the logistic model
• Examine the performance of the model to predict the observed data
• Communicate logistic regression results with confidence

Classroom delivery coupled with hands-on applications via case examples using regression template files and interactive sessions.