Design & Mean Comparisons
Plan a fair experiment, identify the experimental unit, and choose comparisons that answer the scientific question.
Topics¶
Identify what independently receives treatment before counting replicates. Randomization, replication, and blocking protect different parts of the comparison.
Choose the test from the sampling design and variance assumptions. Report the size and uncertainty of the difference, not only a p-value.
For paired data, analyze within-pair differences. Randomization tests must reproduce the assignments allowed by the actual design.
The global test asks whether all treatment means agree. A rejection does not identify which pairs differ.
State the comparison of interest first, then account for the family of comparisons being tested.