Module 05 Activity

Scenario

Choose and run the right test, and report it without overclaiming.

What you build

A test choice justified by four questions, and a results table with intervals.

Steps

1. State H0 as a specific number and H1 including its sidedness, before looking at the data.
2. Answer the four questions — outcome type, group count, paired or independent, assumptions — and pick the scipy call from them.
3. Fix alpha in advance, count how many tests you will run, and say what you will do about multiplicity.
4. Pass equal_var=False to any two-sample t-test and read .confidence_interval() before the p-value.
5. Assemble every planned test into one DataFrame with the estimate, the interval and the p-value, stating the sign convention.
6. Write the findings with estimate, interval, p-value and n, avoiding 'significant' as a substitute for a magnitude.

Evidence to hand in

• H0 and H1, with sidedness, dated before the analysis.
• The four answers and the scipy call they select.
• The number of tests run and your multiplicity decision.
• The results table with its stated sign convention.
• The findings paragraph.

Review checklist

• The alternative's sidedness was fixed before the data was seen.
• equal_var=False is passed to every two-sample t-test.
• The number of tests run is reported.
• Every result carries its confidence interval.
• No non-significant result is described as 'no effect'.

Mark complete