DataTerminology.comNEW
Replacing missing values with the median of observed values in the feature.
DESCRIPTION
Median imputation fills missing entries with the feature’s median, providing robustness against outliers compared to mean imputation. It is simple and effective for skewed distributions. Like other single imputation methods, it understates variance and may bias relationships under MAR or MNAR. Use as a baseline, or when distributional robustness is desired. Prefer multiple imputation for inferential accuracy.
ALTERNATIVE DEFINITIONS
scikit-learn
Replace missing values using the median along each column.
Specifies column-wise median replacement strategy.
Penn State Eberly College of Science
Median imputation replaces missing values with the median of the observed values.
Emphasizes robustness to outliers compared to mean.
RELATED TERMS
KNN Imputation Listwise Deletion Missing Data MCAR MAR MNAR
COMPETING TERMS
Mean Imputation Multiple Imputation
RELATIONS
- Broader: Imputation
KNOWLEDGE AREAS
Data Science AI Data Quality MLOps
TARGET AUDIENCE
Data Scientist AI Engineer MLOps Engineer Data Steward Analytics Engineer
EXTERNAL SOURCES
- Wikipedia - Imputation (statistics)
- scikit-learn - SimpleImputer API docs
- UCLA Statistical Consulting - Overview of missing data handling
- Penn State Eberly College of Science - Lesson describing median imputation
TOP QUESTIONS
When is Median Imputation preferable to Mean Imputation?
Use it for skewed distributions or when outliers are present, as median is more robust than mean.
Category: Requirements
What are common mistakes with Median Imputation?
Applying it without assessing distribution shape or ignoring variance reduction can mislead downstream analysis.
Category: Mistakes
What risks does Median Imputation carry?
It understates variability and can bias correlations under MAR/MNAR mechanisms, affecting inferences.
Category: Risks
How to apply Median Imputation in scikit-learn?
Use SimpleImputer(strategy=‘median’) in a Pipeline, fit on training data, and transform consistently.
Category: HowTo
Is Median Imputation suitable for time series?
Often no; prefer time-aware interpolation or model-based approaches that respect temporal dependencies.
Category: Opinion
What is a hot topic around Median Imputation?
Multiple imputation and autoencoder-based methods better capture uncertainty and complex relationships.
Category: Topic
Does Median Imputation preserve the median?
Yes, but like mean imputation, it reduces variance and can distort covariances.
Category: Misunderstanding
How to validate Median Imputation effectiveness?
Compare performance and inference metrics with alternative methods under cross-validation and sensitivity checks.
Category: Validation
What alternatives exist to Median Imputation?
Mean imputation, KNN imputation, model-based imputation, and multiple imputation.
Category: Alternatives