Module 07 Activity

Scenario

Someone has asked whether two things are related. This activity answers that honestly.

What you build

A relationship analysis with the scatter plotted, confounders checked, and claims limited to what observational data supports.

Steps

1. Plot the scatter before computing anything. Describe the shape in words.
2. Compute the correlation and r-squared, and say how much vertical spread survives a fitted line.
3. Colour the points by every grouping variable you have. Report any relationship that disappears within groups.
4. Write four candidate claims about the relationship and mark which the data supports.
5. Measure the overplotting. If it is substantial, choose a fix and say which, including what it costs.
6. Write the finding as a sentence that would survive a statistician reading it.

Evidence to hand in

• The scatter and your description of its shape.
• Correlation, r-squared and the residual spread.
• The confounder check with any within-group collapse.
• Four claims with the two the data does not support marked.
• The overplotting measurement and your chosen fix.
• The final finding sentence.

Review checklist

• The scatter was plotted before the coefficient was computed.
• r-squared is reported wherever a line is drawn.
• Every available grouping variable was checked as a confounder.
• No claim asserts causation or the result of an intervention.
• Any overplotting fix is named, with what it costs.

Mark complete