GuidedStats assists users with statistical analyses through guided workflows. It automatically verifies assumptions and provides actionable suggestions. It is integrated into notebook coding environment, and exchanges dataset, model and results with the coding environment.
Paper link: Guided Statistical Workflows with Interactive Explanations and Assumption Checking
Tips: you can create a new conda environment to avoid any package conflicts.(Currently fixing errors in jupyter notebooks and jupyter lab>=4.0).
conda create -n test_env jupyter-packaging python=3.11 "jupyterlab<4.0"
First, install the package in editable mode
python -m pip install -e .
And enable widet frontend
# link your development version of the extension with JupyterLab
jupyter labextension develop . --overwrite
To build the project
npm install
npm run build
To automatically rebuild the project after editing frontend code
npm run watch
To rebuild after editing python package, you need to restart the kernel.
Testing...
To initialize the UI, first we need to import GuidedStats and load the dataset into a Pandas DataFrame. Then we pass the DataFrame to GuidedStats.
from guidedstats import GuidedStats
import pandas as pd
df = pd.read_csv("california_housing.csv")
gs = GuidedStats(df)
gs
- Many thanks to Will Epperson and Adam Perer for their kind assistance in my first HCI paper and my early exploration of research interests.
- I am currently focused on incorporating more statistical workflows to empirically demonstrate the scalability of our approach, and simplify the code structure as well.
- As software development is not my primary field of study, I recognize that the code implementation has room for improvement 🙏. I greatly welcome and appreciate any advice or suggestions.