Q: What's the difference between Type I and Type II error?
Type I: false positive (see effect when none exists). Type II: false negative (miss real effect). α controls Type I; power controls Type II.
Q: Why use log probabilities?
Multiplying small probabilities causes underflow. Log turns products into sums.
Q: What sample size do you need?
Depends on: baseline rate, minimum detectable effect, α, power.
Q: What if treatment and control have different sizes?
Unequal sizes are fine, but you lose power. Best practice: : ratio.