AI summary by TwinMind
Quantile-Quantile Plots (QQ plots), Clearly Explained!!!
StatQuest with Josh Starmer
6m 55s
Introduction to QQ Plots
- Josh Stormer introduces quantile-quantile plots (QQ plots)
- Prerequisite: understanding quantiles and percentiles
- Purpose: determine if data follows specific distribution (e.g., normal distribution)
Dataset and Research Question
- Dataset: expression measurements in 15 genes
- Main question: Is this data normally distributed?
- QQ plot will help answer this question
Step-by-Step QQ Plot Construction Process
Step 1: Assign Quantiles to Data Points
- Give each data point its own quantile
- Creates 15 quantiles for 15 data points
Step 2: Select Reference Distribution
- Get normal curve (any normal curve works)
- Will compare data quantiles against normal distribution quantiles
Step 3: Create Matching Quantiles
- Add same number of quantiles to reference curve as data (15 lines)
- Divides both data and normal curve into equal-sized groups
- For normal curve: equal probability within each group
- Edge groups wider (lower probability regions)
- Middle groups narrower (higher probability regions)
Step 4: Plot QQ Graph
- Two axes: data quantiles vs. normal distribution quantiles
- Plotting process for each quantile:
- Smallest data quantile at 0.6 → horizontal dotted line
- Corresponding normal quantile at -1.5 → vertical dotted line
- Place dot at intersection
- Second data quantile at 1.1 → horizontal line
- Second normal quantile at -1.2 → vertical line
- Place dot at intersection
- Third data quantile at 1.9 → horizontal line
- Third normal quantile at -0.89 → vertical line
- Continue process for all quantiles
Interpreting QQ Plot Results
- Check how well dots fit straight line
- If normally distributed: most points on straight line
- Current example: "fit not awesome"
- Conclusion: data not well-matched to normal distribution
Comparison with Uniform Distribution
- Same 4-step process applied to uniform distribution
- Step 1: Give each point quantile
- Step 2: Get uniform distribution
- Step 3: Add same number quantiles to uniform distribution
- Step 4: Plot QQ graph
- Results: points much closer to straight line
- Conclusion: uniform distribution better fit than normal
Comparing Two Datasets
- Second dataset much smaller with only 4 data points
- Uses quartiles (quantiles when only 4 divisions)
- Process:
- Determine 4 quartiles for original dataset
- Compare quartiles between two datasets
- Plot using same intersection method
- Draw horizontal/vertical lines for each dataset's quantiles
- Place dots at intersections
- Limitation: few quantiles make definitive conclusions difficult
- Solution: collect more data for more quantiles and better assessment
Conclusion and Call to Action
- QQ plots useful for determining which distribution matches data
- Encourages subscription and comments for future StatQuest topics
