Zulip Chat Archive

Stream: mathlib4

Topic: statistical models


Zixiao Wang (Mar 31 2025 at 00:30):

Hi! I'm exploring the use of Lean 4 for formalizing statistical models, and I was wondering if there’s any existing guidance or libraries for representing parametric, nonparametric, or otherwise predefined statistical models. I tried searching for “parametric models” in the mathlib or community resources but didn’t find much that seemed related.

If anyone has pointers—whether design patterns, relevant definitions, or example developments—I’d really appreciate it. Thanks!

Best,
Zixiao

Aaron Liu (Mar 31 2025 at 00:31):

What exactly do you mean by "statistical model"?

Zixiao Wang (Mar 31 2025 at 00:33):

like a class or structure of parametric model?

Aaron Liu (Mar 31 2025 at 00:37):

We have stuff like docs#ProbabilityTheory.gaussianReal but I still don't know if that's what you're looking for.

Aaron Liu (Mar 31 2025 at 00:40):

This is "a family of probability distributions that has a finite number of parameters" (quote from Wikipedia)

Zixiao Wang (Mar 31 2025 at 00:47):

thanks for the info! yup i saw this, I was looking for a more generic version rather than restricting the distribution to be gaussian

Aaron Liu (Mar 31 2025 at 00:57):

Wdym generic version? What would a generic version look like?

Aaron Liu (Mar 31 2025 at 00:58):

Maybe docs#MeasureTheory.IsProbabilityMeasure, if you drop "gaussian" from "gaussian probability measure" that's what you get

Aaron Liu (Mar 31 2025 at 00:59):

I don't know what you want


Last updated: May 02 2025 at 03:31 UTC