AI summary by TwinMind
In Statistics, Probability is not Likelihood.
StatQuest with Josh Starmer
5m 1s
Introduction & Overview
- Josh Dahmer presents StatQuest episode on probability vs likelihood
- Two closely related concepts that are easily confused
- Visual approach using pictures to explain differences
Probability Explanation
- Uses normal distribution example with mouse weights
- Distribution parameters:
- Mean: 32 grams
- Standard deviation: 2.5
- Range: 24-40 grams
- Probability = area under curve between specified values
- Example: P(mouse weighs 32-34 grams) = 0.29 (29% chance)
- Mathematical notation: P(32 < X < 34 | μ=32, σ=2.5) = 0.29
- Left side of equation changes for different weight ranges
- Right side (distribution parameters) stays fixed
- Probabilities describe areas under fixed distributions
Likelihood Explanation
- Assumes mouse already weighed (data collected)
- Example: mouse weighs 34 grams
- Likelihood = y-axis value at specific point on curve
- Value at 34 grams = 0.12
- Mathematical notation: L(μ=32, σ=2.5 | X=34) = 0.12
- If distribution shifted (mean = 34), likelihood = 0.21
- Right side (measurements) fixed, left side (distribution parameters) modified
- Likelihoods describe y-axis values for fixed data points with movable distributions
Key Differences Summary
- Probabilities: areas under fixed distribution
- Mathematical form: P(data | distribution)
- Likelihoods: y-axis values for fixed data points
- Mathematical form: L(distribution | data)
- Inverse relationship in what's fixed vs variable
Additional Resources
- Reference to StatQuest on maximum likelihood estimator for exponential distribution
- Subscription and support information
- Link to original songs on Bandcamp
