Closed nmdwolf closed 1 year ago
Yes, I agree, it is difficult to know if the quantiles cross each other or if the prediction cross the quantiles.
Furthermore, the message is written as a tautology : "the upper quantile value might be higher than the lower quantile value". This is the opposite that should trigger the warning.
Finally, it appears that the warning comes even when there is no crossing at all, as can be simply verified on the testing set :
(y_pis[:, 0, 0] > y_pis[:, 1, 0]).sum()
I think that this is due to a wrong unfolding of the arguments at this line, the y_pred
should be in last position according to the code in predict
I agree. So there are two problems:
Describe the bug
When trying to use the MapieQuantileRegressor class I get the following UserWarnings:
The upper quantile better be greater than (or equal to) the lower quantile, so the former warning definitely does not make any sense and the latter one is rather vague. What is meant by the additional value?
To Reproduce Steps to reproduce the behavior:
Expected behavior
I know that quantile regressors can suffer from the quantile crossing problem (without any modifications). However, with the current warning it is hard to see whether this is the problem that occurs or something else is going on.
Desktop
Screenshot![image](https://user-images.githubusercontent.com/48948273/210762648-d7df744e-daaa-4319-9d6a-eb8ad85f375b.png)