Academic Statistics · Module 08 of 11

Regression analysis

This module covers simple and multiple linear regression, hierarchical entry, categorical predictors and interactions, regression diagnostics and an introduction to binary logistic regression. Regression is the workhorse of explanatory and predictive doctoral research, and examiners expect thorough diagnostic checking.

Watch the first lesson of every module free. The remaining lessons, exercises and live sessions open when you enrol.

What you will be able to do

  • Fit and interpret simple and multiple linear regression models, including unstandardised and standardised coefficients
  • Check linearity, independence of errors, homoscedasticity, normality of residuals and multicollinearity using VIF
  • Identify influential cases with Cook's distance and leverage
  • Interpret odds ratios from a binary logistic regression

Key terms

  • Regression coefficient
  • Standardised beta
  • Adjusted R-squared
  • Homoscedasticity
  • Multicollinearity
  • Variance inflation factor
  • Cook's distance
  • Odds ratio

Hands-on practice

Run a hierarchical multiple regression on a provided dataset in SPSS or R, carry out full diagnostics and write a results paragraph with a coefficients table.

Common pitfalls this module fixes

  • Checking normality of the raw outcome rather than of the residuals; the assumption concerns the residuals.
  • Using stepwise selection to build a theory-testing model; enter predictors on theoretical grounds, for example hierarchically.
  • Interpreting a logistic regression coefficient as a change in probability; exponentiate it to an odds ratio and interpret it as a change in odds.