Statistical Significance
Domain: Evaluation
🎯 Motivation
Ensuring evaluation results are not due to random chance. With small samples, results may be misleading.
📋 Concrete Example
Test on 10 questions: 8/10 correct (80%). Is it really 80% accurate, or just lucky? Test on 1000 questions: 750/1000 (75%). Now the result is statistically significant.
🔗 Analogy
A coin flip - 7 heads out of 10 flips. Is the coin biased? Maybe not - small sample. 700 heads out of 1000? Probably biased. Statistical significance requires enough samples.