Math::Regressionizer — README

Math::Regressionizer

Raku package with a class that allows pipeline-like specification and execution of regression workflows.

Extensive guide is given in the Jupyter notebook "Rapid-specification-of-regression-workflows.ipynb" and the corresponding Markdown document "Rapid-specification-of-regression-workflows.md".

Features summary

The class Math::Regressionizer facilitates rapid specifications of regressions workflows.
- To quickly specify:
  - data rescaling and summary
  - regression computations
  - outliers finding
  - conditional Cumulative Distribution Functions (CDFs) reconstruction
  - plotting of data, fits, residual errors, outliers, CDFs
Math::Regressionizer works with datasets, numerical arrays, lists of numbers, and lists of numeric pairs.

Details and arguments

The curves computed with Quantile Regression are called regression quantiles.
Math::Regressionizer has three regression methods:
- quantile-regression
- quantile-regression-fit
- least-squares-fit
The regression quantiles computed with the methods quantile-regression and quantile-regression_fit correspond to probabilities specified with the argument :probs(:@probabilities).
The methodquantile-regression computes fits using a B-spline functions basis.
- The basis is specified with the arguments :@knots and :order(:$interpolation-order).
- :order(:$interpolation-order) is 3 by default.
The methods quantile-regession-fit and least-squares-fit use lists of basis functions to fit with specified with the argument :funcs(:@basis).

Workflows flowchart

The following flowchart summarizes the workflows that are supported by Regressionizer:

i_raw_githubusercontent_com_antononcube_python_regressionizer_main_docs_img_quantile_regression_workflow_extended_jpg

Usage examples

Load packages:

use Math::Regressionizer;
use Data::Generators;

Generate random data:

srand(0);
my @x = 0, 0 + 1/300 ... 2;
my @y = @x.map({ sin(2 * pi * $_) }) <<+>> random-variate(NormalDistribution.new(0, 0.4), @x.elems);
my @data = @x Z @y;

Compute quantile regression for probabilities [0.2, 0.5, 0.8] and make the corresponding plot:

#% html
my $obj = 
  Math::Regressionizer.new(@data)
  .quantile-regression(knots => 8, probs => [0.2, 0.5, 0.8])
  .plot(title => "B-splines fit", template => "dark", format => 'html')

Remark If the code cell above is executed in a Jupyter notebook with a Raku kernel then the plot is shown. "Math::Regressionizer" uses the package "JavaScript::Google::Charts" for its plots.

Export the plot obtained above: