Stat Quest: Statistical Power

Intro

• Hosted by Josh Starburns
• Assumes familiarity with p-values and normal distribution

Key Concepts

Distributions

• Example: Mice weights on two diets (special vs. normal)
• Distributions visually distinct with minimal overlap

P-value Example

• Small sample set comparison
• Small p-value (0.0004) leads to null hypothesis rejection
• High probability of consistently getting small p-values
• Occasional large p-values failing to reject null hypothesis

Definition of Power

• Power: Probability of correctly rejecting null hypothesis
• Large power = high probability of obtaining small p-value
• No difference between diets means power concept doesn't apply

Special Diet with Overlap

• Small difference in distributions with overlap
• Small sample comparison leads to higher p-value (0.34)
• Low probability of rejecting null hypothesis
• Larger sample increases likelihood of small p-values

Increasing Power

• More measurements increase power
• Power analysis determines needed sample size

Summary

• Power: Probability to correctly reject null hypothesis
• More power with distinct distributions and sufficient sample size
• Power analysis aids in determining sample size for adequate power

Closing

• Invitation to subscribe and support Stat Quest
• Links for Patreon, channel membership, merchandise, and donations available