ML 3.6Lesson outlineLesson 6 of 7 in Module 3
Regression Diagnostics and Robust Regression
Read residual plots to find what a linear model gets wrong, tell an outlier from a point that controls the fit, and choose a loss that one bad row cannot dominate.
Video in production.
Lesson outline: Regression Diagnostics and Robust Regression
In this lesson10 parts
- 01Residuals against fitted values
- 02The curve: add a feature or transform the target
- 03The fan: heteroscedasticity
- 04An outlier, and a row with leverage
- 05One mistyped number
- 06Losses one row cannot dominate
- 07RANSAC
- 08What to do about each
- 09Code
- 10Recap, your challenge, and what comes next
Builds on
This lesson stands on its own. These go deeper into what it uses.
Key terms: Regression Diagnostics and Robust Regression
The words this lesson introduces, each in one line. The module’s glossary collects them all, and the AI glossary has every term across the courses.
- fitted value
- The model's prediction for a training row.
- residual plot
- Residuals plotted against fitted values.
- heteroscedasticity
- Residuals whose spread changes with the fitted value.
- outlier
- A row far from the model's prediction.
- leverage
- How far a row's features are from the other rows', which sets how much it can turn the fit.
- hat matrix
- The matrix that turns
yinto the fitted values; its diagonal is the leverage. - influence
- How much a row moves the fit.
- Cook's distance
- How much the fitted values change when one row is left out.
- absolute loss
- The size of the residual.
- Huber loss
- Squared inside a threshold and linear outside it.
- robust regression
- Fitting with a loss that a few bad rows can not dominate.
- median (reminder, Maths 4.3)
- The value that minimises the absolute loss.
- RANSAC
- Fitting on random minimal samples and keeping the one most rows agree with.
- consensus set
- The rows within the threshold of the chosen fit.
Quiz: Regression Diagnostics and Robust Regression
Your first pick on each question is the one that counts. Getting one wrong here is how the lesson sticks.