As mentioned in Small things wrt "Using numerical and categorical variables together", the use of “significantly” (e.g. the gradient boosting tree is significantly better than the logistic regression) is not very precise.
We should either:
- not use “significantly” stay vague and say the gradient boosting tree performs better than the logistic regression
- be more precise and explain what we mean by significantly and why we can say it
An example of “significantly” I removed recently : we were saying than the logistic regression is significantly better than the majority class predicter based on a train-test split.