Introduction to Logistic Regression
Many investigations often have data where there are only two possible categories of response - for example a surgical procedure may result or may not result in a complication, a crop may or may not become infested with pests, a patient's symptoms may or may not involve a sore throat.
This course will explain the use of logistic regression for studying associations between binary outcomes like those mentioned, and possible explanatory factors. The emphasis will be on practical application and interpretation rather than theory. A large component of the course will be PC-based practical work on user-friendly statistics packages such as Minitab, R, SAS, SPSS and Stata.
Scientists and technologists who already have some statistical training but whose knowledge is lacking in the area of regression methods for binary response data. Prior attendance on A Review of Basic Statistics and Regression Analysis: A Hands-on Approach, or equivalent knowledge, is required for this course.
You will be introduced to the increasingly widely used modelling technique of logistic regression for analysing binary response data, and learn how to fit and interpret such models.
- Binary data: examples, ungrouped and grouped data
- Why is logistic regression required? Benefits
- Logistic regression model for binary response data: link functions, odds, concept of likelihood for model fitting
- Quantifying effects of explanatory variables: odds ratios, hypothesis testing and confidence intervals
- Model comparisons and selection strategies
- Assessing the goodness-of-fit of a logistic regression model: goodness-of-fit tests and residuals
- Use of logistic regression models for prediction, ROC curves and inverse prediction (e.g. estimating an ED50)
- Computational problems and potential remedies
- Presentation of results from a logistic regression.
Dates: 15-16 May 2014
Duration: 2 days
Discounts: An Academic discount is available for this course.
Last updated: 16 October, 2013