Implementation incorrect

[j-hartshorn]

The ale plot should not be anchored at the midpoint of the bucket. This is because the ALE value represents the average change in response from the bottom of the bucket to the top.

As a simple example:

If each bucket, the average local effect in each bucket, and the observation weight (usually number of observations) are as follows:

bucket	average_local_effect	weight
[0, 1]	22	15
(1, 2]	36	25
(2, 3]	-10	35
(3, 4]	-41	25

Then the (non-centered) ALE will be:

bucket_edge	ALE
0	0
1	22
2	58
3	48
4	7

This is because 22 is the change in prediction between 0 and 1, 36 is the change between 1 and 2, and so on.

We then center with the constant -1.45 (from average_local_effect @ weight / sum(weight)) to get the final ALE:

bucket_edge	ALE
0	-1.45
1	20.55
2	56.55
3	46.55
4	5.55

See the original implementation here.

[akuhnregnier]

Hi @j-hartshorn , thanks for bringing this up. I actually noticed this as well - I think I fixed this here: https://github.com/akuhnregnier/ALEPython (specifically in this commit).

Does my updated implementation match the original implementation in this regard? It would be great to get a second opinion on this before I invest the time to prepare a pull request (many of my changes are probably not very relevant in general).