Statista - Advanced Statistical Analysis Package

Overview

Statista is a comprehensive Python package for statistical analysis, focusing on probability distributions, extreme value analysis, and sensitivity analysis. It provides robust tools for researchers, engineers, and data scientists working with statistical models, particularly in hydrology, climate science, and risk assessment.

Current release info

Name	Downloads	Version	Platforms

conda-forge feedstock

Conda-forge feedstock

Installation

Conda (Recommended)

conda install -c conda-forge statista

PyPI

pip install statista

Development Version

pip install git+https://github.com/serapeum-org/statista

Main Features

Statistical Distributions

• Probability Distributions: GEV, Gumbel, Normal, Exponential, and more
• Multi-Distribution Fitting: Fit all distributions at once and select the best fit
• Parameter Estimation Methods: Maximum Likelihood (ML), L-moments, Method of Moments (MOM)
• Goodness-of-fit Tests: Kolmogorov-Smirnov, Chi-square
• Truncated Distributions: Focus analysis on values above a threshold

Extreme Value Analysis

• Return Period Calculation: Estimate extreme events for different return periods
• Confidence Intervals: Calculate confidence bounds using various methods
• Plotting Positions: Weibull, Gringorten, and other empirical distribution functions

Sensitivity Analysis

• One-at-a-time (OAT): Analyze parameter sensitivity individually
• Sobol Visualization: Visualize parameter interactions and importance

Statistical Tools

• Descriptive Statistics: Comprehensive statistical descriptors
• Time Series Analysis: Auto-correlation and other time series tools
• Visualization: Publication-quality plots for statistical analysis

Quick Start

Single Distribution

import numpy as np
from statista.distributions import Distributions

# Load your data
data = np.loadtxt("examples/data/time_series2.txt")

# Create a distribution object and fit parameters
dist = Distributions("Gumbel", data=data)
params = dist.fit_model(method="lmoments", test=False)
print(params.loc, params.scale)

# Calculate PDF and CDF
pdf = dist.pdf(plot_figure=True)
cdf, _, _ = dist.cdf(plot_figure=True)

# Goodness-of-fit tests
ks_stat, ks_pvalue = dist.ks()
chi_stat, chi_pvalue = dist.chisquare()

Multi-Distribution Fitting

from statista.distributions import Distributions

# Fit all distributions and find the best one
dist = Distributions(data=data)
best_name, best_info = dist.best_fit()
print(f"Best: {best_name}")
print(f"Parameters: {best_info['parameters']}")

# Or fit all and inspect results
results = dist.fit()
for name, info in results.items():
    print(f"{name}: KS p-value={info['ks'][1]:.4f}")

Extreme Value Analysis

from statista.distributions import Distributions, PlottingPosition

# Fit a GEV distribution using L-moments
gev_dist = Distributions("GEV", data=data)
params = gev_dist.fit_model(method="lmoments")

# Calculate non-exceedance probabilities
cdf_weibul = PlottingPosition.weibul(data)

# Calculate confidence intervals
lower_bound, upper_bound, fig, ax = gev_dist.confidence_interval(
    plot_figure=True
)

For more examples and detailed documentation, visit Statista Documentation

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

This project is licensed under the MIT License - see the LICENSE file for details.

Citation

If you use Statista in your research, please cite it as:

Farrag, M. (2023). Statista: A Python package for statistical analysis, extreme value analysis, and sensitivity analysis.
https://github.com/serapeum-org/statista

BibTeX:

@software{statista2023,
  author = {Farrag, Mostafa},
  title = {Statista: A Python package for statistical analysis, extreme value analysis, and sensitivity analysis},
  url = {https://github.com/serapeum-org/statista},
  year = {2023}
}