How can I calculate the 95%CI for F1 score, Matthews correlation coefficient and fraction correct ((TN+TP)/(TP+TN+FP+FN)) when I evaluating a perfomance of binominal diagnostic test?
Solved – How to calculate 95%CI for F1 score and Matthews correlation coefficient
confidence intervalcorrelationmachine learningmathematical-statistics
Related Question
- Solved – Matthews correlation coefficient – how much is a good classifier
- Solved – Measuring Statistical Significance of Binary Classification using Matthews Correlation Coefficient
- Solved – How to interpret Matthews correlation coefficient (MCC)
- Solved – Confidence interval on point biserial correlation coefficient
- Solved – Matthews correlation coefficient with multi-class
Best Answer
F1, MCC and other such metrics can be thought of as proportions, since they are in [0,1]. The binomial proportion confidence interval can be estimated with (using normal approximation)