Types of Missing Data: MCAR, MAR, and MNAR
What you'll learn: You'll understand the three fundamental mechanisms that cause missing data and why distinguishing between them matters for choosing the right handling strategy.
Why Missing Data Mechanisms Matter
Before you can properly handle missing values in your dataset, you need to understand why they're missing. The pattern of missingness dramatically affects which imputation or handling strategies will work without introducing bias into your model.
The Three Missing Data Mechanisms
Missing Completely At Random (MCAR)
Data is MCAR when the probability of a value being missing is completely independent of any data—observed or unobserved.
Analogy: A lab technician randomly spills coffee on survey forms, erasing random entries. The missingness has nothing to do with the actual values.
Missing At Random (MAR)
Data is MAR when the probability of missingness depends on observed data, but not on the missing values themselves.
Analogy: Younger survey respondents skip income questions more often than older ones. If you know someone's age, you can predict the likelihood their income is missing—but the actual income value doesn't influence whether it's missing.
Missing Not At Random (MNAR)
Data is MNAR when the probability of missingness depends on the unobserved value itself.
Analogy: High-income earners specifically refuse to report their income because it's high. The missing value itself causes the missingness—the hardest scenario to handle properly.
Why This Matters
MCAR allows simple deletion without bias. MAR permits sophisticated imputation using other features. MNAR requires domain knowledge or specialized modeling, as the missingness contains information about the hidden values.
Key Takeaway: Understanding whether your missing data is MCAR, MAR, or MNAR guides your feature engineering strategy and determines whether simple approaches will introduce bias or preserve data integrity.