Mailman 3 May 2021 - SciPy-Dev

numpy/npy_math.h vs C/C++ stdlib math.h/<cmath>
by Evgeni Burovski May 9, 2021

May 9, 2021

Hi, Prompted by Sturla's comment, https://github.com/scipy/scipy/pull/13931#discussion_r619728364, suggesting to change a call of `npy_log` into an `std::log`, here's a question: what is the recommendation going forward for the math functions (log, exp, trig functions etc) in scipy C/C++ code: do we import them from `numpy/npy_math.h` or do we rely on the standard library? My main concern is more exotic platforms/toolchains and related edge cases--- do we rely on numpy or the standard library vendors? `npy_math.h` still often needs to be included for e.g. numpy data types, so it's not a question of avoiding an extra #include. Cheers, Evgeni

3 2

Update: Inverse of Log CDF of Normal Distribution
by Steppi, Albert May 4, 2021

May 4, 2021

Dear SciPy developers, I mentioned last week that I have an implementation of the inverse of the log CDF of the normal distribution that I'd like to contribute to scipy. scipy.special currently contains the functions ndtr, ndtri, and log_ndtr, for the CDF, the inverse CDF, and the log of the CDF of the normal distribution respectively. I propose adding a function ndtri_exp for the inverse of the log CDF. Earlier I submitted the issue gh-13923 regarding this. Today I've submitted the PR gh-13979 with an enhancement that adds ndtri_exp as a ufunc to scipy.special. I think this could be a valuable function to add. It has proven useful in my group's research and was specifically requested in gh-11465 so is likely to be useful to others. I understand that you are all very busy and may not have time to attend to this PR immediately. I appreciate all the work you do; I've been a scipy user for over a decade and have found it an immensely valuable resource. I hope someone else may find this ndtri_exp implementation valuable. Thanks, Albert -- Albert Steppi III, PhD Scientific Software Developer Laboratory of Systems Pharmacology Harvard Medical School

2 2

Proposal to add `bootstrap_ci` function for estimating confidence intervals of n-sample statistics
by Matt Haberland May 1, 2021

May 1, 2021

Dear SciPy Developers, One of the items in the detailed SciPy roadmap is: > Where appropriate, include confidence intervals for the statistic in the results of any statistical test. gh-13371 proposes a function, `bootstrap_ci`, that accepts two arguments, `data` and `statistic`, and (by default) estimates a 95% confidence interval for the true value of the statistic. https://github.com/scipy/scipy/pull/13371 Features: - Implements "percentile", "reverse percentile/basic", and "BCa" bootstrap - Works on n-sample statistics (for now, "BCa" is only available for one-sample statistics) - Works along the provided `axis` of n-dimensional data The PR does not add a confidence interval (or a method for calculating the confidence interval) to the information returned by existing statistical tests, but the function `bootstrap_ci` can easily be used with existing statistical tests by providing a lightly-wrapped version of the test as the `statistic` argument (see examples in documentation). Adding this functionality to existing tests (and new tests) for convenience is being discussed for a follow-up PR. Please consider joining the discussion about this proposed feature in gh-13371! Thanks, Matt -- Matt Haberland Assistant Professor BioResource and Agricultural Engineering 08A-3K, Cal Poly

1 0