A blank cell is not a method—what should you decide before imputing?
Start here
A blank cell is not a method—what should you decide before imputing?
Missingness has a cause. Deletion, simple imputation, local imputation or model-native handling can each be reasonable, but only when their assumptions match how values became missing.
Building interactive view…
Understand
Build the mental model
Missingness has a cause. Deletion, simple imputation, local imputation or model-native handling can each be reasonable, but only when their assumptions match how values became missing. Profile missingness by feature, subgroup and time. Compare strategies on downstream validation and inspect post-imputation distributions.
Click a stage to inspect what happens, what changes, and what should be checked before moving on.
Stage 1
Profile missingness
For the “Profile missingness” stage, identify the incoming object, the rule applied to it, the state change produced, and the evidence that would reveal a mistake. Technical context for Missing Data Decisions: MCAR, MAR and MNAR describe relationships between missingness and observed/unobserved values. Imputation parameters must be learned inside training folds to avoid leakage.
Practitioner checkpoint: Profile missingness by feature, subgroup and time. Compare strategies on downstream validation and inspect post-imputation distributions.
What happens if…?
Break the assumption deliberately
Impute before splitting the data and identify where information from the test set leaks backward.
Move the control and explain what you expect before reading the visual.
Technical lens
Formalise what the visual is doing
MCAR, MAR and MNAR describe relationships between missingness and observed/unobserved values. Imputation parameters must be learned inside training folds to avoid leakage.
Technical questionUse a tiny case to make the mechanism observable. MCAR, MAR and MNAR describe relationships between missingness and observed/unobserved values. Imputation parameters must be learned inside training folds to avoid leakage. Verify one intermediate quantity, state change or mapping independently; then predict the consequence of this change: Impute before splitting the data and identify where information from the test set leaks backward.
Practitioner lens
Use it responsibly
Profile missingness by feature, subgroup and time. Compare strategies on downstream validation and inspect post-imputation distributions.
Transfer testImputing before train/test splitting.
Worked exploration
Use the visual as an experiment, not decoration
Create a table where age is missing more often in one region. Compare overall missingness with missingness by region. Try deletion and median imputation, then inspect how sample size and the age distribution change.
Technical lens
MCAR, MAR and MNAR describe relationships between missingness and observed/unobserved values. Imputation parameters must be learned inside training folds to avoid leakage.
Practitioner check
Profile missingness by feature, subgroup and time. Compare strategies on downstream validation and inspect post-imputation distributions.
Prediction before interaction
Impute before splitting the data and identify where information from the test set leaks backward.
Exploration walkthrough
Turn the interaction into an evidence trail
Create a table where age is missing more often in one region. Compare overall missingness with missingness by region. Try deletion and median imputation, then inspect how sample size and the age distribution change. Before moving the control, state your prediction. After the visual changes, name the specific state, statistic, boundary or mapping that changed and explain why that change is consistent—or inconsistent—with your prediction.
Record one observable quantity before the interaction and the same quantity afterwards.
Change one factor at a time so the causal effect of the control is inspectable.
Use an edge or failure case to discover where the concept stops behaving as the simple story suggests.
Static orientation diagram for Missing Data Decisions; use the interactive visual above to test how the relationships change.
Reference depth
Open the complete material
The flagship experience is the map. These pages contain the roads.
