Categories → Causal Wizard Concept
Causal Wizard is implemented with Open-Source Python packages.
Causal Wizard is built on top of several very popular and widely used Python packages. There are so many that only some key packages are mentioned here.
For core number crunching, we use NumPy and Pandas, two of the most globally popular scientific and technical computing tools.
For Causal Inference, we are using DoWhy, part of PyWhy. We will also extend the models to include those supported by EconML and possibly some custom models as well. So far, we only support DoubleML from EconML.
If you're an experienced Python programmer, you can also use DoWhy directly, although you may still find Causal Wizard a convenient shortcut.
The DoWhy paradigm of Identification, Estimation and Refutation is followed pretty closely in Causal Wizard, although we've wrapped these steps in more thorough checks and added a lot of error handling and messaging around issues detected. In addition we always perform a broad suite of additional validation checks and other options in DoWhy, such as Bootstrap resampling for confidence intervals and significance testing. In short, we try to make it easy for non-experts to use.
Causal Wizard was built in Django. The user interface HTML/CSS is based on Bootstrap.
We'd also like to call out Cytoscape.js, which we use for causal diagram model rendering and interaction. Other plots are rendered with Plotly.js.