6.3 Roadmap

Looking Backwards

• Statisticians create a model to represent the world
• We saw examples of estimators, which approximate model parameters we’re interested in.
• By itself, an estimate isn’t much good; we need to capture the uncertainty in the estimate.
• We’ve seen two ways to express uncertainty in an estimator: standard errors and confidence intervals.

Today

• We introduce hypothesis testing
  • A hypothesis test also captures uncertainty, but in relation to a specific hypothesis.

Looking Ahead

• We’ll build on the one-sample t-test, to introduce several other statistical tests.
• We’ll see how to choose a test from different alternatives, with an eye on meeting the required assumptions, and maximizing power.