Missing Data Mechanisms & Imputation Bias Lab

Data-centric AI treats the dataset, not the model, as the thing worth engineering — and missing values are the single most common data-quality defect a pipeline inherits. This lab generates a clean synthetic dataset with a known linear relationship, then lets you hide values under one of the three classic missingness mechanisms (MCAR, MAR, MNAR) and repair the gaps with listwise deletion, mean imputation or regression imputation. Every hidden point stays visible in red as ground truth so you can watch, in 3D, exactly how far each repair strategy's estimated slope and mean drift from the truth — the same silent bias that corrupts a model trained on real-world incomplete data.