One day – an introduction to the statistical methods
Four parts of ninety minutes, with the statistics in the foreground
This version of the one-day course keeps the statistical methods in the foreground. The case and the exercise are those of every format, a Multi100 reanalysis of a claim made by the original author (OA) of a paper about how the European Union is framed across its member states, in which you reproduce a result recorded by Multi100, write the estimand behind the claim, work a menu of defensible alternatives and preregister one of them before you run it. What changes is where the time goes. The causal-graph crash course runs in full. The specification menu is taken apart axis by axis, to show what each choice does to an estimate: the outcome scale and family, the form of the predictor, the four estimators and what fixed effects adjust for, the copredictor, the sample and the weights. And the closing part treats the multiverse at length, from where the idea came from to why hundreds of estimates on one dataset are not hundreds of tests.
| What happens | Length | |
|---|---|---|
| Before the day | Install R and an IDE, download the workspace, read the claim, and fit one linear model. | 1 to 3 hours, once |
| Part 1 – The case and the vocabulary | The paper and the three verdicts reached on it, the five analysts who reanalysed its claim, and what ‘replication’ names; then you get running. | 90 minutes |
| Part 2 – The claim and its estimand | The paper in full, estimands, and the causal-graph crash course; you write the estimand and finish the graph. | 90 minutes |
| Part 3 – Reproduction and the specification menu | You reproduce the recorded result, then work through the axes of the menu that set the question and the model, one at a time, and put an exploratory dot on the class chart. | 90 minutes |
| Part 4 – Multiverse and preregistration | Reading the curve, what it establishes and what it cannot, and the analysis you preregister before it runs. | 90 minutes |
| After the day | Every model on the menu output by output, and the rest of the methods strand. | as you like |