sarlib 0.0.3

Statistical Analysis and Regression Library

SARlib: Statistical Analysis and Regression Library

This library provides tools for statistical analysis, regression modeling, sample size analysis, and visualization. It includes OLS and SAR models, as well as utilities for data preprocessing and plotting.

Installation

SARlib can be installed via PyPI:

pip install sarlib

Alternatively, you can install it manually by downloading the source code. In that case, make sure you have the following dependencies installed:

• numpy
• matplotlib
• statsmodels
• scikit-learn
• scipy
• pyside (only for GUI)

Usage

Import the module in your Python script:

import sarlib

Or copy the code into your project and import the classes/functions as needed.

Main Components

• fix_data(x, y): Standardizes and cleans input data.

• show_scatter(x, y, ...): Visualizes predictors vs. response.

• OLS: Ordinary Least Squares regression with permutation-based significance and power analysis.

• SAR: Statistical Analysis Regression with PAC-Bayes, Vapnik, and IGP bounds.

• SampleSizeAnalysis: Analyzes the effect of sample size on model performance and statistics.

Example Workflow

1. Import packages and prepare your data as numpy arrays:

   from sarlib import SAR, OLS, SampleSizeAnalysis, show_scatter
   import numpy as np
   x = np.random.randn(100, 3)  # predictors
   y = np.random.randn(100)     # response
   

2. Visualize data:

   show_scatter(x, y)
   

3. Fit SAR model:

   model_sar = SAR(n_realiz=100, norm='epsins', alpha=0.05)
   stats_sar = model_sar.fit(x, y, verbose=True)
   

4. Compare with an OLS model:

   model_ols = OLS(n_realiz=100, alpha=0.05)
   stats_ols = model_ols.fit(x, y, verbose=True)
   

5. Analyze sample size effect:

   analysis = SampleSizeAnalysis(model_sar, x, y, steps=7)
   analysis.plot_loss()
   analysis.plot_pvalue()
   analysis.plot_coef()
   

Function/Class Documentation

All functions and classes are documented with docstrings. Please refer to the code for parameter details and usage.

License & Author

Author: Sipba Group, UGR, https://sipba.ugr.es/ License: GPL Version 3