first version of joss paper, scipy implmentation geometric distribution

paper/paper.bib

@@ -0,0 +1,51 @@
+@article{marsaglia2004evaluating,
+  title   = {Evaluating the anderson-darling distribution},
+  author  = {Marsaglia, George and Marsaglia, John},
+  journal = {Journal of statistical software},
+  volume  = {9},
+  pages   = {1--5},
+  year    = {2004}
+}
+
+@book{walck1996hand,
+  title     = {Hand-book on statistical distributions for experimentalists},
+  author    = {Walck, Christian and others},
+  year      = {1996},
+  publisher = {Stockholms universitet}
+}
+
+@article{george2011estimation,
+  title   = {Estimation of parameters of Johnson's system of distributions},
+  author  = {George, Florence and Ramachandran, KM},
+  journal = {Journal of Modern Applied Statistical Methods},
+  volume  = {10},
+  pages   = {494--504},
+  year    = {2011}
+}
+
+@article{sinclair1988approximations,
+  title     = {Approximations to the distribution function of the anderson—darling test statistic},
+  author    = {Sinclair, CD and Spurr, BD},
+  journal   = {Journal of the American Statistical Association},
+  volume    = {83},
+  number    = {404},
+  pages     = {1190--1191},
+  year      = {1988},
+  publisher = {Taylor \& Francis}
+}
+
+@book{mclaughlin2001compendium,
+  title     = {A compendium of common probability distributions},
+  author    = {McLaughlin, Michael P},
+  year      = {2001},
+  publisher = {Michael P. McLaughlin}
+}
+
+@article{lewis1961distribution,
+  title     = {Distribution of the Anderson-Darling statistic},
+  author    = {Lewis, Peter AW},
+  journal   = {The Annals of Mathematical Statistics},
+  pages     = {1118--1124},
+  year      = {1961},
+  publisher = {JSTOR}
+}

paper/paper.md

@@ -0,0 +1,73 @@
+---
+title: "Phitter: A Python Library for Probability Distribution Fitting and Analysis"
+tags:
+    - Python
+    - Statistics
+    - probability distributions
+    - data analysis
+    - machine learning
+    - simulation
+    - monte carlo
+authors:
+    - name: Sebastián José Herrera Monterrosa
+      orcid: 0009-0002-2766-642X
+      affiliation: 1
+affiliations:
+    - name: Pontificia Universidad Javeriana
+      index: 1
+date: 26 March 2024
+bibliography: paper.bib
+---
+
+# Summary
+
+Phitter is a Python library designed to analyze datasets and determine the best analytical probability distributions that represent them. It provides a comprehensive suite of tools for fitting and analyzing over 80 probability distributions, both continuous and discrete. Phitter implements three goodness-of-fit tests and offers interactive visualizations to aid in the analysis process. For each selected probability distribution, Phitter provides a standard modeling guide along with detailed spreadsheets that outline the methodology for using the chosen distribution in various fields such as data science, operations research, and artificial intelligence.
+
+# Statement of Need
+
+In the fields of data science, statistics, and machine learning, understanding the underlying probability distributions of datasets is crucial for accurate modeling and prediction. However, identifying the most appropriate distribution for a given dataset can be a complex and time-consuming task. Phitter addresses this need by providing a user-friendly, efficient, and comprehensive tool for probability distribution fitting and analysis.
+
+Phitter stands out from existing tools by offering:
+
+1. A wide range of over 80 probability distributions, including both continuous and discrete options.
+2. Implementation of multiple goodness-of-fit tests (Chi-Square, Kolmogorov-Smirnov, and Anderson-Darling).
+3. Interactive visualizations for better understanding and interpretation of results.
+4. Accelerated fitting capabilities for large datasets (over 100K samples).
+5. Detailed modeling guides and spreadsheets for practical application in various fields.
+
+# Features and Functionality
+
+Phitter offers a range of features designed to streamline the process of probability distribution analysis:
+
+-   **Flexible Fitting**: Users can fit both continuous and discrete distributions to their data.
+-   **Customizable Analysis**: Options to specify the number of bins, confidence level, and distributions to fit.
+-   **Parallel Processing**: Support for multi-threaded fitting to improve performance.
+-   **Comprehensive Output**: Detailed summaries of fitted distributions, including parameters, test statistics, and rankings.
+-   **Visualization Tools**: Functions to plot histograms, PDFs, ECDFs, and Q-Q plots for visual analysis.
+-   **Distribution Utilities**: Methods to work with individual distributions, including CDF, PDF, PPF, and sampling functions.
+
+# Implementation and Usage
+
+Phitter is implemented in Python and is available via PyPI. It requires Python 3.9 or higher. The library can be easily installed using pip:
+
+```
+pip install phitter
+```
+
+Basic usage involves creating a `PHITTER` object with a dataset and calling the `fit()` method:
+
+```python
+import phitter
+
+data = [...]  # Your dataset
+phi = phitter.PHITTER(data)
+phi.fit()
+```
+
+More advanced usage allows for customization of fitting parameters and specific distribution analysis.
+
+# Conclusion
+
+Phitter provides researchers, data scientists, and statisticians with a powerful tool for probability distribution analysis. By offering a comprehensive set of distributions, multiple goodness-of-fit tests, and interactive visualizations, Phitter simplifies the process of identifying and working with probability distributions in various data-driven fields.
+
+# References

phitter/__init__.py

@@ -1,4 +1,4 @@
-__version__ = "0.0.8"
+__version__ = "0.7.1"
 
 from .main import PHITTER
 from phitter import continuous

phitter/discrete/discrete_distributions/geometric.py

@@ -43,16 +43,16 @@ def cdf(self, x: int | numpy.ndarray) -> float | numpy.ndarray:
         """
         Cumulative distribution function
         """
-        result = 1 - (1 - self.p) ** numpy.floor(x)
-        # result = scipy.stats.geom.cdf(x, self.p)
+        # result = 1 - (1 - self.p) ** numpy.floor(x)
+        result = scipy.stats.geom.cdf(x, self.p)
         return result
 
     def pmf(self, x: int | numpy.ndarray) -> float | numpy.ndarray:
         """
         Probability mass function
         """
-        result = self.p * (1 - self.p) ** (x - 1)
-        # result = scipy.stats.geom.pmf(x, self.p)
+        # result = self.p * (1 - self.p) ** (x - 1)
+        result = scipy.stats.geom.pmf(x, self.p)
         return result
 
     def ppf(self, u: float | numpy.ndarray) -> float | numpy.ndarray:
@@ -67,7 +67,7 @@ def sample(self, n: int, seed: int | None = None) -> numpy.ndarray:
         Sample of n elements of ditribution
         """
         if seed:
-            numpy.random.seed(0)
+            numpy.random.seed(seed)
         return self.ppf(numpy.random.rand(n))
 
     def non_central_moments(self, k: int) -> float | None:

pyproject.toml

@@ -4,7 +4,7 @@ build-backend = "setuptools.build_meta"
 
 [project]
 name = "phitter"
-version = "0.0.8"
+version = "0.7.1"
 description = "Find the best probability distribution for your dataset"
 authors = [{name = "Sebastián José Herrera Monterrosa", email = "[email protected]"}]
 readme = "README.md"
@@ -36,7 +36,8 @@ dependencies = [
     "scipy>=1.1.0",
     "plotly>=5.14.0",
     "kaleido>=0.2.1",
-    "matplotlib>=3.3"
+    "matplotlib>=3.3",
+    "pandas>=1.5.0"
 ]
 
 [project.urls]

tests/phitter_local/distributions/test_distributions_continuous.ipynb

@@ -2,7 +2,7 @@
  "cells": [
   {
    "cell_type": "code",
-   "execution_count": 5,
+   "execution_count": 1,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -11,7 +11,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 2,
    "metadata": {},
    "outputs": [],
    "source": [