Skip to content
ML Academy

Introduced on day 6

MAR (missing at random)

Missingness that depends only on observed variables; conditioning or imputing on those variables removes the bias.

In detail

Example: avg_fare missing more often on intl-long flights, but unrelated to the fare within each market. Pooled complete-case totals are biased; within-market analysis or market-conditional imputation is not.