Doc. Add an article on statistics in Rust

Provide "Notes on statistics in Rust" with corresponding references in
README.

diff --git a/README b/README
index 1f34732a7d36c98691b16071da7d67f5b89e7293..eae7a4ed3cd543c5f55f4b8133154276e37f980f 100644 (file)
--- a/README
+++ b/README
@@ -9,10 +9,8 @@ Decide an available distribution, e.g. the log normal, and apply the
 quantiles from that on a ranked priority list. The output will be the
 priority in percentages per item in the list.
 
-Read more:
-
- * Wikipedia, [Ranking](https://en.wikipedia.org/wiki/Ranking)
- * Wikipedia, [Quantile](https://en.wikipedia.org/wiki/Quantile)
+Read more in [Notes on statistics in Rust](doc/stats.md) on the
+details of implementation.
 
 
 Installation

diff --git a/doc/stats.md b/doc/stats.md
new file mode 100644 (file)
index 0000000..bd31703
--- /dev/null
+++ b/doc/stats.md
@@ -0,0 +1,37 @@
+---
+title: Notes on statistics in Rust
+author: Gustav Eek
+date: 24 jan 2023
+---
+
+<!--
+Copyright (C) 2023 Gustav Eek <gustav.eek@fripost.org>
+SPDX-License-Identifier: GPL-3.0-or-later
+-->
+
+In blogs, [Scientific ...][g:r], *rust-sci.github.io*, provides a
+brief overview; Filipe, [Using rust ...][r:u], *esciencecenter.nl*,
+2021, compares Rust with C/Fortran approaches, praises *nalgebra*,
+iterators and the borrow checker, as well as Rust's built in
+documentation, testing, and parallelization features.
+
+[g:r]: https://rust-sci.github.io/
+[e:u]: https://blog.esciencecenter.nl/using-rust-for-scientific-numerical-applications-learning-from-past-experiences-798665d9f9f0
+
+In summary, the problem is to implement [quantile][w:q] based
+quantified [ranking][w:r] from [log-normal
+distributions][w:l]. Possible approaches includes (1) implement
+something from scratch or (2) rely on some library. Candidates are
+
+ 1. [statrs][g:s] -- A wrapper around some C# library
+ 2. [russell stat][g:c] -- standalone one man project
+
+[w:q]: https://en.wikipedia.org/wiki/Ranking
+[w:q]: https://en.wikipedia.org/wiki/Quantile
+[w:l]: https://en.wikipedia.org/wiki/Log-normal_distribution
+[g:s]: https://github.com/boxtown/statrs
+[g:c]: https://github.com/cpmech/russell
+
+The former supports quantiles in *inverse_cdf*, implemented as binary
+search. The latter only the CDF.
+