Introduction
Link to video transcript in a Word file
Most health outcomes are multifactorial, meaning that there are multiple factors that influence whether a given outcome will occur, and these other risk factors can introduce confounding that distorts our primary analysis. While stratification is very useful for adjusting for confound by one or two confounders at a time, it is not an efficient way to adjust for multiple confounding factors in a single analysis. In this module we will first extend our discussion of simple linear regression to introduce you to multiple linear regression in which we evaluate multiple independent variables when looking at a continuously distributed dependent variable (health outcome.) We will then introduce multiple logistic regression analysis as a tool for evaluating associations between exposures and dichotomous outcomes. We will draw on examples from the public health literature to discuss the interpretation of results from multiple linear and logistic regression models. You will also gain skills in using R to conduct and interpret these analyses using a public health data set.
Essential Questions
- How can we simultaneously take into account multiple factors that might affect a health outcome?
- How can we calculate an adjusted measure of effect, controlling for other factors, through statistical analysis?
- What analysis should be done if the outcome is dichotomous instead of continuous? How I perform multivariable regression analyses in R?
Learning Objectives
After completing this section, you will be able to:
- Explain the principle of statistical adjustment to a lay audience
- Identify when linear regression or logistic regression models should be used to control for confounding
- Interpret regression coefficients (slopes) in multiple linear regression analysis
- Interpret adjusted odds ratios in multiple logistic regression
- Explain the relationship between confidence intervals for slopes or odds ratios and p- values for slopes or odds ratio in regression models
- Conduct a multiple variable regression analysis using the R statistical package
- Interpret multiple variable regression output generated from the R statistical package