@thomasjpfan @lucyleeow Do you remember if there is a reason for choosing name instead of estimator_name in this display?

I think the initial API used "estimator_name", because I thought "name" was too generic. Moving forward, I am okay with using "name". I will be happy to review the PR.

Moving forward, I am okay with using "name". I will be happy to review the PR.

Basically, it is to use estimator_name everywhere instead :)

Now that we have "from_predictions", there is not always an estimator. So "name" is better now?

with from_predictions, you can pass estimator_name as well. If nothing is provided, we default to None.

For example, PrecisionRecallDisplay.from_predictions accepts name and sets the estimator_name=name. I think that inconsistency is weird.

I guess a PR to make CalibrationDisplay to be more like PrecisionRecallDisplay is okay for the time being. It's better to make the change now since CalibrationDisplay has not been officially released yet.

PrecisionRecallDisplay.from_predictions does not set estimator_name:

name only allows for overwritting estimator_name.

However, CalibrationDisplay.plot is setting self.name and this is weird:

I indeed corrected this in the PR.

Ah yes, in that case CalibrationDisplay.plot needs to be fixed.

To be specific, the issue I was thinking about was the user API:

display = PrecisionRecallDisplay.from_predictions(..., name="Model one")

# The current way to get the name of the curve
display.estimator_name

# Is this better?
display.name