Discrimination Threshold plot explanation

[robmcd]

Describe the issue
Hi, I have been trying to use the discrimination threshold plot as per below

from yellowbrick.classifier import PrecisionRecallCurve, DiscriminationThreshold

# Precision-Recall Plot
pr_curve = PrecisionRecallCurve(model)
pr_curve.fit(X_train, y_train)
pr_curve.score(X_test, y_test)
pr_curve.show()

# Threshold Plot
threshold_plot = DiscriminationThreshold(model)
threshold_plot.fit(X_train, y_train)
threshold_plot.score(X_test, y_test)
threshold_plot.show()

However, I'm confused by the results. Perhaps there's a gap in my understanding.

My precision recall plot shows values of precision ranging from 0.2 to 0.6, with a "jump" to 1 at zero recall..

My discrimination threshold plot shows "scores" for precision ranging from 0.5 up to 1 (talking about the line, not the band). The bands are very narrow and only widen from threshold 0.9+

On the precision recall chart a recall of 0.2 relates to a precision of ~0.4. However on the discrimination threshold chart a recall of 0.2 relates to a precision score of ~0.9?

Why does the recall vs. precision figures I see on the PR chart not match the precision vs recall scores on the discrimination threshold chart?

Apologies for my ignorance. Could you help me understand?

Also, I don't understand how the "score" for precision can start at 0.5 (for a discrimination threshold value of 0) given that in the precision-recall curve we saw values for precision as low as 0.2 (presumably for some low probability threshold). Is the varying discrimination threshold not the same as the varying probabilities used to generate the precision recall curve? Why is there nowhere on the discrimination threshold plot showing a precision of 0.2?

@DistrictDataLabs/team-oz-maintainers