StatTools

Statistical learning and inference algorithms implemented in pure Python (version 3.6 or later).

Installation

The latest version of StatTools can be installed directly after cloning from GitHub:

git clone https://github.com/artemmavrin/StatTools.git
cd StatTools
make install

Moreover, StatTools is on the Python Package Index (PyPI), so a recent version of it can be installed with the pip utility:

pip install stattools

Dependencies

• NumPy
• SciPy
• pandas
• Matplotlib

Examples

Regression

• Simple linear regression for fitting a line through a scatter plot
• Ridge regression
• Elastic net regularization (including LASSO and ridge regression as special cases)
• Fitting a polynomial curve to a scatter plot
• Various scatterplot smoothers applied to a sine curve with Gaussian noise

Classification

• Logistic regression for breast cancer diagnosis
• Multiclass logistic regression for handwritten digit recognition

Unsupervised Learning

• K-means clustering for grouping unlabelled data together
• Estimation of Gaussian mixture models
• Principal component analysis applied to handwritten digits
• Kernel density estimation for histogram smoothing

Non-Parametric Statistics

• The bootstrap (ordinary and Bayesian) and the jackknife for standard error estimation
• Bootstrap confidence intervals
• Exact and Monte Carlo permutation tests
• The Kaplan-Meier survivor function estimator

Ensemble Methods

• Using bagging to improve logistic regression accuracy

Data Visualization

• Plotting lines and function curves
• Drawing empirical distribution functions
• Drawing quantile-quantile (QQ) plots

Simulation

• Simulating sample paths of Poisson processes
• Simulating sample paths of Itô diffusions (for example, Brownian motion)