Using unregistered scales

MPL Users,

Using [1] as a template, I built a ProbabilityScale[2] which I hope one day will be in the statsmodels library. It still needs some work, but it’s mostly there.

I have, what I hope is a basic question:

Do I have to register the scale in order to use it, or is there an axes method to feed it the ProbScale class directly?

I notice that there is an axes.xaxis._set_scale method – but there appears to be no “public” API way of using an unregistered scale.

If it is the case that there’s no sanctioned way of applying an unregistered scale to an Axes object, when/where in a module is best to register it (e.g., on import, locally within a the function/method that needs it)?

Cheers,

-paul

[1] http://matplotlib.org/examples/api/custom_scale_example.html?highlight=codex%20set_major_locator

Here’s that second link for the scale: https://gist.github.com/phobson/3cc3550ce1efcc299142

···

On Mon, Oct 6, 2014 at 1:27 PM, Paul Hobson <pmhobson@…287…> wrote:

MPL Users,

Using [1] as a template, I built a ProbabilityScale[2] which I hope one day will be in the statsmodels library. It still needs some work, but it’s mostly there.

I have, what I hope is a basic question:

Do I have to register the scale in order to use it, or is there an axes method to feed it the ProbScale class directly?

I notice that there is an axes.xaxis._set_scale method – but there appears to be no “public” API way of using an unregistered scale.

If it is the case that there’s no sanctioned way of applying an unregistered scale to an Axes object, when/where in a module is best to register it (e.g., on import, locally within a the function/method that needs it)?

Cheers,

-paul

[1] http://matplotlib.org/examples/api/custom_scale_example.html?highlight=codex%20set_major_locator

Hi Paul,

I built a ProbabilityScale[2] which I hope one day will be in the statsmodels library.

This just made me think that back in April I was also playing with matplotlib scales for probability distribution.

It's actually a completely different idea, because I was doing a "logit" scale to get a good visualization of *tails* in a cumulated distribution plot. So I'm jumping on your thread in case anyone can give me some feedback on this idea. For example, I have no clue on how common this kind of plot is (and how useful it can be!!).

best,
Pierre

(and going back to your original question, I notice that I was indeed registering the scale...)

···

Le 06/10/2014 22:27, Paul Hobson a écrit :