Regression for Statisticians
Regression models how a response variable depends on one or more predictors. Linear regression fits a line by least squares and is checked with residual diagnostics, while generalised linear models extend the idea to binary outcomes with logistic regression and to counts with Poisson regression. Coefficients are read as effects, with uncertainty.
Stage: Modelling · Part of the Statistician roadmap · Last reviewed 19 November 2026
What to learn in Regression
- Fit a multiple linear regression and interpret each coefficient
- Check linearity, constant variance and normality of residuals with plots
- Detect multicollinearity using variance inflation factors
- Fit a logistic regression and read coefficients as odds ratios
- Compare models using adjusted R-squared, AIC and cross-validation
Regression interview questions
What are the main assumptions of linear regression?
The relationship is linear in the parameters, errors are independent, errors have constant variance, and errors are roughly normal for inference in small samples. Residual plots help check these. Violations can be addressed with transformations, robust errors or a different model.
How do you interpret a logistic regression coefficient?
A coefficient is the change in the log-odds of the outcome for a one-unit increase in the predictor, holding others fixed. Exponentiating it gives an odds ratio. A coefficient of 0.7 means the odds are multiplied by about 2, not the probability.
Where Regression fits in the Statistician roadmap
Regression belongs to the Modelling stage (intermediate level). Relate variables.
Other units in this stage
Back to the full Statistician roadmap