Free AI Website MakerEDEF
Python package that computes feature importance for machine learning models based on Euler decompositions of explained fit, in contrast to the common focus on predictions. The method decomposes model fit, is exactly additive, and is (much) faster than many alternative approaches. Additional details on GitHub and PyPI.
Installation: pip install edef
TreeIG
Python package that computes integrated gradients feature importance for tree-based machine learning models. Although regular gradients are not well defined for tree-based models, integrated gradients reduce exactly to the sum of discrete prediction steps along the IG feature path. Additional details on GitHub and PyPI.
Installation: pip install treeig
MATLAB code for various statistical computations. Includes regression code and many statistical distribution functions. Many of these functions were made available in MATLAB toolboxes at later dates.
Regressions
Run OLS regressions with a variety of standard error corrections. (Requires Distributions.)
LAD
Run least absolute deviations regression. (Requires Distributions)
Logit
Run maximum likelihood logistic regressions. (Requires Distributions.)
Factor Analysis
Perform maximum likelihood factor analysis. (Requires Distributions, Statistics, Strings/IO.)
Distributions
Evaluate statistical distribution functions.
Statistics
Compute and display sample statistics. (Requires Distributions, Strings/IO).
Strings/IO
Utilities for string handling and IO.
Date/Time
Utilities for handling dates.
Utility
Some utilities that don't fit anywhere else.
MATLAB code that performs financial calculations used in some of my published papers.
Implied Volatility
Perform efficient implied volatility estimation when prices are observed with measurement errors. See "Errors in implied volatility estimation.” (Requires the following packages from my web pages: Options, Distributions, Utility, Strings/IO. The last three are in the Statistical Computations section.)
Options
Compute Black–Scholes option prices, derivatives, implied volatilities, and optimal implied volatility averages. (Requires the following Statistical Computations packages from my web pages: Distributions, Utility.)
Power Utility
Efficiently find optimal portfolio weights for a power-utility investor by avoiding local optima and discontinuities. See “Numerical solution of the static portfolio problem for power utility investors.”
Gauss code for maximum likelihood estimation of many different GARCH models used in some of my published papers
GARCH
Estimate standard GARCH models using numerical maximum likelihood.
AGARCH
Estimate GARCH models from "All in the family: Nesting symmetric and asymmetric GARCH models."
NNIGN
Estimate GARCH models from "No news is good news: An Asymmetric model of changing volatility in stock returns."
You can install the Python packages using Pip or from GitHub. The packages contain internal documentation.
For the other software, after downloading the archives, decompress them (see below), and install the functions somewhere on the search path. Do this either by placing the functions in a directory on the current path or by updating the path.
The functions are documented using the standard MATLAB/GAUSS help system. In MATLAB command mode, type "help fn" to get basic help for the function fn. Based on this internal documentation, there should be no difficulty in determining calling conventions or interpretation of results.
The major archives also contain a ReadMe.m file that outlines the basic purpose of the functions in the archive.
All archives have been compressed using Zip and require the appropriate decompression software. Zip utilities are part of many operating systems.
All functions and programs are supplied without any guarantees or support. Although I have used the main features of the functions successfully, some of the functions contain features I have rarely if ever used.
Several of the functions contain known or suspected errors in secondary features. These errors are flagged in the functions with "ERROR: ... ".
If you find additional errors in the functions, please notify me, and I may try to fix them.
If you use these programs, please clearly cite their origin.
Free AI Website Maker