open-risk/concentrationmetrics

A python library for the computation of various concentration, inequality and diversity indices

Intro

concentrationMetrics is a python library for the computation of various concentration, diversification and inequality indices. The library implements the computation of all well known indexes of inequality and concentration.

You can use concentrationMetrics to

• access an exhaustive collection of concentration, inequality and diversity indexes and metrics
• perform file input/output in both json and csv formats
• compute indexes with confidence intervals via bootstraping
• visualize using matplotlib

Key Information

• Author: Open Risk
• License: MIT
• Mathematical Documentation: Open Risk Manual
• Code Documentation: Read The Docs
• Development website: Github
• Discussions: Open Risk Commons

NB: concentrationMetrics is still in active development. If you encounter issues please raise them in our github repository

Full Description

List of Implemented Indexes

An overview of the implemented metrics (indexes) and their relationships is available at the Open Risk Manual

The below list provides more specific documentation URL's for each one of the implement indexes:

• Atkinson Index
• Berger-Parker Index
• Concentration Ratio
• Ellison-Glaeser Index
• Gini Index
• Theil Index
• Hannah-Kay Index
• Hoover Index
• Herfindahl-Hirschman Index and related indexes such as Simpson and Inverse Simpson
• Shannon Index
• Generalized Entropy Index (Renyi)
• Kolm Index

Support and Training

The Open Risk Academy has free courses demonstrating the use of the library: Open Risk Academy

Examples

Comparing two indexes across a range of input portfolio data

Calculating industrial and geographic concentrations

Contributions

• Check the TODO list for immediate TODO items
• Check the Roadmap for the bigger picture

concentrationmetrics's Issues

Categorical data processing

Margalef index implemented as first example, more to follow

Error instead of termination

If input criteria are not of the expected form (such as that the input is not a numpy array), the object throws a message and then terminates the execution. Neither is standard or desirable. Instead, the object should throw an exception (probably AttributeError) and nothing else.

graph_example: AttributeError: module 'networkx' has no attribute 'write_yaml'

Document installation

This looks like a fine library! I think it can help me a lot.

Alas, I don't know how to get it. Your documentation doesn't state how to get it, it just starts with import concentrationMetrics as cl. But neither concentrationMetrics nor concentrationmetrics (package names should be all lower-case as per PEP8!) exist on PyPI. If you want users to download a script you should tell them.

tests: Warning in install.packages : package ‘IC2’ is not available for this version of R

tests: Failed: [pytest] section in setup.cfg files is no longer supported, change to [tool:pytest] instead.

ellison_glaeser_example: TypeError: list indices must be integers or slices

Add Lorenz curve functionality

Integrate Lorenz curve calculation / plotting along with Gini index

Gini formula looks wrong (but works)

Looking for the formula here http://www.openriskmanual.org/wiki/Gini_Index

It seem that this code is wrong (but it works perfectly)

(1.0 - 2.0 * np.multiply(i, weights).sum())/n + 1.0

I think the right code would be:

(2.0 * np.multiply(i/n, weights).sum()) - (n - 1.0)/n

or following the other formula

(n + 1.0)/n - (2.0 * np.multiply((n + 1 - i)/n, weights).sum())

Those two codes that match the formulas have a problem, they get the results as negative numbers

Streamline normalization / scaling of all indexes

Provide "standard" choice by default, offer additional options via parameter

.

.