Open kan-qi opened 5 years ago
Currently the p-values for accuracy improvements are calculated by t-test:
Since the distribution of test statistic is not normal, we need to use non-parametric bootstrapping test to calculate the p-value
Example code can be found: https://stats.stackexchange.com/questions/136661/using-bootstrap-under-h0-to-perform-a-test-for-the-difference-of-two-means-repl
Currently the p-values for accuracy improvements are calculated by t-test:
Since the distribution of test statistic is not normal, we need to use non-parametric bootstrapping test to calculate the p-value
Example code can be found: https://stats.stackexchange.com/questions/136661/using-bootstrap-under-h0-to-perform-a-test-for-the-difference-of-two-means-repl