Spurious Correlations
4 days ago
- #correlation
- #data-science
- #statistics
- Spurious correlations are random connections between unrelated variables.
- Examples include the popularity of memes correlating with the number of air traffic controllers in Montana.
- Other examples link GMO use in corn to pirate attacks globally.
- The article discusses data dredging, where large datasets yield random correlations.
- Lack of causal connection means these correlations are coincidental, not meaningful.
- Observations are not independent, as sequential years may show trends.
- Y-axes on graphs are often truncated, making correlations appear stronger than they are.
- Confounding variables, like global events, can create misleading correlations.
- Outliers in data can disproportionately influence correlation results.
- Low sample sizes (n) can make correlations unreliable despite high p-values.