<h1 align="center"> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/IMAGES/LOGO/plotastic_logo_3.png" width="700px" alt="logo"> </h1> <h1 align="center">

</h1>

plotastic: Bridging Plotting and Statistics

📦 Installation

Install from PyPi:

pip install plotastic

Install from GitHub: (experimental, check CHANGELOG.md)

pip install git+https://github.com/markur4/plotastic.git

Requirements

• Python >= 3.11 (not tested with earlier versions)
• pandas == 1.5.3 (pingouin needs this)
• seaborn <= 0.12.2 (later versions reworked hue)

📷 Example Gallery

<details><summary> <b><i> (click to unfold) </i></b> </summary> <blockquote> <hr> <h1 align="center"> <hr> 🐁 Click on Images for Code! 🐁 <hr>

<a href=https://github.com/markur4/plotastic/blob/main/EXAMPLES/qpcr.ipynb> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/qpcr1.png" alt="qpcr1"> </a> <a href=https://github.com/markur4/plotastic/blob/main/EXAMPLES/fmri.ipynb> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/fmri2.png" width="500px" alt="fmri2"> </a> <a href="https://github.com/markur4/plotastic/blob/main/EXAMPLES/attention.ipynb"> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/attention1.png" width="250px" alt="attention"> </a> <a href=https://github.com/markur4/plotastic/blob/main/EXAMPLES/tips.ipynb> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/tips1.png" width="350px" alt="tips1"> </a> <a href=https://github.com/markur4/plotastic/blob/main/EXAMPLES/iris.ipynb> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/iris1.png" width="400px" alt="iris1"> </a> <a href="https://github.com/markur4/plotastic/blob/main/EXAMPLES/cars.ipynb"> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/cars1.png" alt="cars1"> </a>
<a href="https://github.com/markur4/plotastic/blob/main/EXAMPLES/diamonds.ipynb"> <img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/diamonds1.png" alt="diamonds1">
<img src="https://raw.githubusercontent.com/markur4/plotastic/main/EXAMPLES/diamonds2.png" alt="diamonds2"> </a>

</h1> </blockquote> </details>

🧑‍🏫 About plotastic

<details><summary> 🤔<b><i> Summary </i></b> </summary> <blockquote> <hr>

plotastic addresses the challenges of transitioning from exploratory data analysis to hypothesis testing in Python's data science ecosystem. Bridging the gap between seaborn and pingouin, this library offers a unified environment for plotting and statistical analysis. It simplifies the workflow with a user-friendly syntax and seamless integration with familiar seaborn parameters (y, x, hue, row, col). Inspired by seaborn's consistency, plotastic utilizes a DataAnalysis object to intelligently pass parameters to pingouin statistical functions. The library systematically groups the data according to the needs of statistical tests and plots, conducts visualisation, analyses and supports extensive customization options. In essence, plotastic establishes a protocol for configuring statical analyses through plotting parameters. This approach streamlines the process, translating seaborn parameters into statistical terms, providing researchers and data scientists with a cohesive and user-friendly solution in python.!

Workflow:

1. 🧮 Import & Prepare your pandas DataFrame
   • We require a long-format pandas dataframe with categorical columns
   • If it works with seaborn, it works with plotastic!
2. 🔀 Make a DataAnalysis Object
   • DataAnalysis(DataFrame, dims={x, y, hue, row, col})
   • Check for empty data groups, differing samplesizes, NaN-count, etc. automatically
3. ✅ Explore Data
   • Check Data integrity, unequal samplesizes, empty groups, etc.
   • Quick preliminary plotting with e.g. DataAnalysis.catplot()
4. 🔨 Adapt Data
   • Categorize multiple columns at once
   • Transform dependent variable
   • Each step warns you, if you introduced NaNs without knowledge!
   • etc.
5. ✨ Perform Statistical Tests ✨
   • Check Normality, Homoscedasticity, Sphericity
   • Perform Omnibus tests (ANOVA, RMANOVA, Kruskal-Wallis, Friedman)
   • Perform PostHoc tests (Tukey, Dunn, Wilcoxon, etc.) based on pg.pairwise_tests()
6. 📊 Plot figure
   • Use pre-defined and optimized multi-layered plots with one line (e.g. strip over box)!
   • Annotate statistical results (p-values as *, **, ***, etc.) with full control over which data to include or exclude!
7. 💿 Save all results at once!
   • One DataAnalysis object holds:
     • One DataFrame in self.data
     • One Figure in self.fig, self.axes
     • Multiple statistical results: self.results
   • Use DataAnalysis.save_statistics() to save all results to different sheets collected in one .xlsx filesheet per test

</blockquote> </details> <details><summary> 📊<b><i> Translating Plots into Statistics!</i> </b> </summary> <blockquote> <hr>

In Principle:

• Categorical data is separable into seaborn's categorization parameters: x, y, hue, row, col. We call those "dimensions".
• These dimensions are assigned to statistical terms:
  • y is the dependent variable (DV)
  • x and hue are independent variables (IV) and are treated as within/between factors (categorical variables)
  • row and col are grouping variables (categorical variables)
  • A subject may be specified for within/paired study designs (categorical variable)
• For each level of row or col (or for each combination of row- and col levels), statistical tests will be performed with regards to the two-factors x and hue

Example with ANOVA:

• Imagine this example data:
  • Each day you measure the tip of a group of people.
  • For each tip, you note down the day, gender, age-group and whether they smoke or not.
  • Hence, this data has 4 categorical dimensions, each with 2 or more levels:
    • day: 4 levels (monday, tuesday, wednesday, Thursday)
    • gender: 2 levels (male, female)
    • smoker: 2 levels (yes, no)
    • age-group: 2 levels (young, old)
• Each category is assigned to a place of a plot, and when calling statistical tests, we assign them to statistical terms (in comments):

  •   # dims is short for dimensions
      dims = dict(          # STATISTICAL TERM:
          y = "tip",        # y-axis, dependent variable
          x = "day",        # x-axis, independent variable (within-subject factor)
          hue = "gender",   # color,  independent variable (within-subject factor)
          col = "smoker",   # axes,   grouping variable
          row = "age-group" # axes,   grouping variable
      )
    

• We perform statistical testing groupwise:
  • For each level-combinations of smoker and age-group, a two-way ANOVA will be performed (with day and gender as between factors for each datagroup):
    • 1st ANOVA assesses datapoints where smoker=yes AND age-group=young
    • 2nd ANOVA assesses datapoints where smoker=yes AND age-group=old
    • 3rd ANOVA assesses datapoints where smoker=no AND age-group=young
    • 4th ANOVA assesses datapoints where smoker=no AND age-group=old
  • Three-way ANOVAs are not possible (yet), since that would require setting e.g. col as the third factor, or implementing another dimension (e.g. hue2).

</blockquote> </details> <details><summary> <b>❗️<i> Disclaimer about Statistics </i></b> </summary> <blockquote> <hr>

This software was inspired by ...

• ... Intuitive Biostatistics - Fourth Edition (2017); Harvey Motulsky
• ... Introduction to Statistical Learning with applications in Python - First Edition (2023); Gareth James, Daniela Witten, Trevor Hastie, Robert Tibshirani, Jonathan Taylor
• ... talking to other scientists struggling with statistics

✅ plotastic can help you with...

• ... gaining some practical experience when learning statistics
• ... quickly gain statistical implications about your data without switching to another software
• ... making first steps towards a full statistical analysis
• ... plotting publication grade figures (check statistics results with other software)
• ... publication grade statistical analysis IF you really know what you're doing OR you have back-checked your results by a professional statistician
• ... quickly test data transformations (log)

🚫 plotastic can NOT ...

• ... replace a professional statistician
• ... teac