Plotly Express hovermode with arbitrary column

Question

I see that the hovermode attribute in layout has options for x or y, but is it possible to use an arbitrary dataframe column? instead?

For example, I'm plotting precision-recall curves. The x-axis is recall, and the y-axis is precision. The independent variable is a detection threshold value (with range of np.linspace(0,1.0,101)) with a column in my dataframe called threshold. When I hover over a precision-recall point, what I'm interested in is points on other curves with the same detection threshold value. So can I instead hover on this column?

Braunite

So you have multiple precision-recall curves in this context? – Caterina

Yes, one for each class label, and then a micro-averaged curve as well – Braunite

Answer 1

No, there’s no feature for that at the moment. You could consider a second set of “iso-threshold” curves to see this kind of thing perhaps?

Caterina

I have been considering iso-threshold curves, although the downsides are clutter (one of the reasons I wanted to use interactive plots in the first place) and needing to deselect/select two matching legend items for each. – Braunite