Regression
These materials introduce regression as a way to translate an observed engineering relationship into a quantitative prediction. We continue to use the course gas-turbine emissions dataset, beginning with a single, interpretable question before adding the complications of multiple predictors and model validation.
We highly recommend working with Google Colab. It lets you run the notebooks in the cloud without local setup.
1. Simple linear regression: weather and NOx
Estimated time: 45–60 minutes
Use ambient temperature to predict NOx concentration, fit a least-squares line, and distinguish a statistically significant association from a causal claim or a complete engineering model.
Dataset source
The underlying measurements come from the Gas Turbine CO and NOx Emissions dataset at the UCI Machine Learning Repository. The course version adds a campaign_year field and documented missing values for later data-pipeline exercises.