Keyword collision in scipy.stats.levy_stable.interval

If you check the methods documentation of the Levy-stable continuous random variable you will see interval(alpha, alpha, beta, loc=0, scale=1). The problem is here that the first parameter for interval itself is named alpha and in the case of the Levy-stable distribution the first shape parameter is also called alpha. This makes it really problematic if you use keyword arguments when calling interval for a user-selected distribution. The only way out of it is renaming the the first parameter of interval or maybe less breaking is renaming the shape parameters of levy_stable.

[mdhaber]

Rather than changing the shape parameter name n of those seven distributions, I’d suggest changing the name of the parameter n of the scipy.stats.rv_continuous.moment method to either

• moment, for consistency with the parameter name moment in scipy.stats.moment, or
• order, which would also be a reasonable name.

We could also consider changing the name of parameter alpha of rv_continuous.interval to confidence_level as in stats.binomtest, but maybe this name only makes sense when the interval in question is thought of as a confidence interval.

Since these are the first arguments of the methods, they may be less likely to be specified as keywords. If that’s true, perhaps this would be less disruptive to users than changing the shape parameter names of some very common distributions. (Full disclosure, I’m thinking of what headache it would be to change all the documentation and tutorials of these distributions to reflect a shape parameter name change - especially n. I imagine changing a parameter name of the methods would be much easier as they are probably not often used as keywords in SciPy code.)

Shall I propose this on the mailing list?

(Happy to push this forward a bit by e.g. drafting the test in gh-13490, but if it gets very messy, I might not see this through to completion any time soon. It’s looking like the cost/benefit of addressing this issue will be unfavorable, and it’s been waiting quite patiently for the past few years, so perhaps it doesn’t need to be a top priority.)

[mdhaber]

Closed by gh-13490. Well, the keyword collision is fixed, but the documentation issue still persists. If we fix gh-10328 (actually render distribution-specific documentation), that will be fixed, too.

Update: gh-15619 will fix the documentation and we’ll let that close this.