Interactive one-page demo that shows, in real time, how different missing-data mechanisms could bias regression.
Move the sliders, watch β̂ diverge from the truth, and build intuition about MCAR vs MNAR in seconds.
- Percent-missing slider plus presets for MCAR, MNAR-by-X, MNAR-by-Y,
X ∨ Y, Flatten and Steepen. - Live read-outs of βtrue, β̂, and a colour-coded bias gauge.
- One-click resample to illustrate sampling variability.
- Single HTML file
With AI! 'vibe coding' - that is, giving the idea to an AI (LLM) and asking it to implement it. I used a combination of LLMs (GPT, claude, gemini), and worked iteratively untill a reasonably useful tool was created - sometimes the LLMs make mistakes (errors, unwanted features etc). It was then hosted on github.
MIT License - see the LICENSE file for details.