It does this by evaluating the prediction problems of The 2 versions over a particular interval. The examination checks the null speculation that the two versions have the very same overall performance on normal, against the alternative that they don't. If the examination statistic exceeds a critical value, we reject the null hypothesis, indicating