Chi-squared test

KDD

Part of Correlation Analysis.

χ2=i=1nj=1m(cijeij)2eij\chi^{2}=\sum_{i=1}^{n} \sum_{j=1}^{m} \frac{\left(c_{i j}-e_{i j}\right)^{2}}{e_{i j}}
  • zero → no Correlation
  • larger value → more Correlation
  • cells with very differenct actual count compared to expected count contribute the most