Week 7 · Multiple linear regression. Practice
Three worked examples end to end: hedonic land prices, a rice production function, and returns to fertiliser in a field experiment
The final week applies everything: each example goes from problem setting and model formulation through exploratory analysis, estimation, interpretation, checking linearity and re-specifying the model, to a conclusion. The third example shows how regression is used to read the results of a randomised experiment.
Slides
| Material | WS 2024/25 | WS 2023/24 |
|---|---|---|
| 🖥 Interactive slides | Week 07. Multiple Linear Regression. Practice | not rendered (sources were revised for 2024/25) |
| 📄 PDF from Ilias | Week 07. Multiple Linear Regression. Practice.pdf · 55 pages · added 19 Nov 2024 | Week 07. Multiple Linear Regression. Practice.pdf · 56 pages · added 30 Jan 2024, file replaced 14 Oct 2024 |
| 💻 Source | mk68-07-slides.qmd, which includes mlr-example-land-price.qmd, mlr-example-production-function.qmd and mlr-rct-example-duflo2008.qmd |
mk68-07-slides.qmd |
What the deck covers
- Causal interpretation of the MLR: multiple regression controls only for the factors included in the model; external ceteris paribus is assumed; the regression assumptions restated.
- Example 1, a hedonic land prices model: problem setting, regression model formulation, exploratory analysis (summary statistics, regional differences, distributions, scatter plots), estimating the regression, interpreting the results, checking linearity, re-specifying the model and re-checking, and comparing the two models.
- Example 2, a production function: problem setting, the regression model and ex-ante hypotheses about the coefficients, exploratory analysis including the effect of the BIMAS programme, estimating the regression and interpreting it coefficient by coefficient, checking linearity, moving to a log specification, interpreting the log model, and conclusions.
- Example 3, returns to fertiliser in Kenya: objective and results of the field experiment, whether ceteris paribus is ensured in the RCT, the authors’ regression solution, and checks and balances.
- Regression in Excel: a pointer for those who do not use R.
Data used
MinnLand(Minnesota farmland sales) from the alr4 R package, for the hedonic land prices model.RiceFarms(Indonesian rice farms) from the plm R package, for the production function.- Tables of the published field experiment, as images in
materials/img.
Where the course goes next
← Week 6 · Simple and multiple regression. Bias. Linearity · All slide decks · Reading list