2 - 9-9-2026

Type I error - we rejected the null hypothesis, we found an effect but there should've been no effect.

Type II error - we have not rejected the null hypothesis - we've found there's no effect when there should've been no effect

Power: the chance your test catches an effect that is really there

Power goes up with:

• A larger true effect
• A larger sample size $n$

Not detected is not the same as "not there"

Assumptions of a t-test:

• Independence (every row is its own piece of information). Broken when data is influenced by other data.
• The estimate is roughly normal, which is not the data being normal.
• The groups share a spread, or use Welch.
• No outliers.