Knowledge MaxxingMissing-value imputation
Missing-value imputation fills in a reasonable estimate for data points that are missing, so a model can use a row instead of discarding it entirely.
Prerequisite: Feature engineering
What matters
The imputation method chosen, such as using an average or a more informed estimate, can itself introduce bias, so how missingness is handled deserves as much care as any other modeling decision.
Example
A dataset with some missing income values might fill each gap with the median income of similar customers, rather than leaving it blank or dropping the row.
Common misconception
Imputation is not simply about avoiding dropped rows; a poorly chosen imputation value can quietly distort the patterns a model learns.
Recall question
Why is choosing how to impute missing values a modeling decision, not just a technical fix?