Metadata-Version: 2.1
Name: minimal-stats
Version: 1.0.0
Summary: A small package for statistical distributions
Home-page: https://github.com/ashishlamsal/minimal-stats.git
Author: ASHISH LAMSAL
Author-email: ashishlamsal.me@gmail.com
License: UNKNOWN
Project-URL: Bug Tracker, https://github.com/ashishlamsal/minimal-stats/issues
Platform: UNKNOWN
Classifier: Programming Language :: Python
Classifier: Programming Language :: Python :: 3.8
Classifier: Programming Language :: Python :: 3.9
Classifier: License :: OSI Approved :: MIT License
Classifier: Operating System :: OS Independent
Requires-Python: >=3.8
Description-Content-Type: text/markdown
License-File: LICENSE

# [minimal-stats][pypi Link]

This is a simple python package for statistical distributions. Currently this package calcuates Binomial and Gaussian distribution.

![build][build]&nbsp;
![Python][Python]&nbsp;
![pypi][pypi]&nbsp;
![wheel][wheel]&nbsp;
![code size][code size]&nbsp;
[![Contributor Covenant][Contributor Covenant]][code_of_conduct.md file]&nbsp;
[![License][License]][License file]&nbsp;

## Installation

```bash
    pip install minimal-stats
```

## Gaussian Distribution

We can directly provide the mean and standard deviation of data (or read data from a file) and add two gaussain distribution.

```python
    >>> from distributions import Gaussian
    >>> g1 = Gaussian(180, 34)
    >>> g1
    g = Gaussian(mean=180, stdev=34)

    >> str(g1)
    'mean 180, standard deviation 34'

    >>> g2 = Gaussian(180, 34)
    >>> g1 + g2
    g = Gaussian(mean=360, stdev=48.08326112068523)
```

Here, we read data from a file, calculate mean, standard deviation and probability density function of gaussian distribution and then see graphical output

```python
    >>> from distributions import Gaussian
    >>> g = Gaussian()
    >>> g.read_data_file(r'\tests\input\numbers.txt')

    >>> g.calculate_mean()
    78.0909090909091
    >>> g.calculate_stdev()
    92.87459776004906
    >>> g.pdf(5)
    0.0031515485379333356
    >>> g.plot_histogram_pdf()
```

![gaussian image][gaussian]

## Binomial Distribution

We can directly provide the n and p of data (or read from file as before) and calculate mean, standard deviation and probability mass function of binomial distribution.

```python
    >>> from distributions import Binomial
    >>> b = Binomial(0.15, 60)
    >>> b
    b = Binomial(p=0.15, n=60)

    >>> b.calculate_mean()
    9.0
    >>> b.calculate_stdev()
    2.765863337187866
    >>> b.pmf(7)
    0.11985659270959788

    >>> a = Binomial(0.15, 50)
    >>> a + b
    b = Binomial(p=0.15, n=110)
```

Here, we read data from a file, `b.replace_stats_with_data()calculate mean, standard deviation, n and p of Binomial distribution and then we plot bar graph of pmf.

```python
    >>> from distributions import Binomial
    >>> b.read_data_file(r'\tests\input\numbers_binomial.txt')
    >>> b.replace_stats_with_data()
    >>> str(b)
    'mean 8.0, standard deviation 1.7541160386140584, p 0.6153846153846154, n 13'
    >>> b.plot_bar_pmf()
```

![binomial image][binomial]

## Contribution

We appreciate feedback and contribution to this repo! Before you get started, please see the following:

- [Contribution Guidelines][CONTRIBUTING.md file]
- [Code of Conduct Guidelines][code_of_conduct.md file]

## License

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

[gaussian]: https://raw.githubusercontent.com/ashishlamsal/minimal-stats/main/sample/gaussian_distribution.png
[binomial]: https://raw.githubusercontent.com/ashishlamsal/minimal-stats/main/sample/binomial_distribution.png

[pypi link]: https://pypi.org/project/minimal-stats/
[build]: https://img.shields.io/github/workflow/status/ashishlamsal/minimal-stats/Package?style=flat-square
[Python]: https://img.shields.io/badge/-Python-3776AB?style=flat-square&logo=python&logoColor=ffffff
[pypi]: https://img.shields.io/pypi/v/minimal-stats?style=flat-square
[wheel]: https://img.shields.io/pypi/wheel/minimal-stats?style=flat-square
[code size]: https://img.shields.io/github/languages/code-size/ashishlamsal/minimal-stats?style=flat-square
[Contributor Covenant]: https://img.shields.io/badge/Contributor%20Covenant-2.1-4baaaa.svg (code_of_conduct.md)
[License]: https://img.shields.io/github/license/ashishlamsal/minimal-stats?style=flat-square (LICENSE)
[License file]: https://github.com/ashishlamsal/minimal-stats/blob/main/LICENSE
[CONTRIBUTING.md file]:  https://github.com/ashishlamsal/minimal-stats/blob/main/CONTRIBUTING.md
[code_of_conduct.md file]: https://github.com/ashishlamsal/minimal-stats/blob/main/code_of_conduct.md

