MNAR Functions¶
Missing Not At Random (MNAR) refers to a missing data mechanism where the probability of missingness depends on the unobserved data.
Formally, let \(X\) be the data matrix and \(M\) the missingness indicator. Then under MNAR:
This implies that the probability of missingness may depend directly on the values that are missing — making the mechanism inherently more complex and harder to address using standard imputation techniques.
In MissMecha, we provide multiple MNAR variants for both numerical and categorical data. These functions help simulate challenging scenarios for evaluating robustness of imputation and modeling pipelines.
Note
MNAR mechanisms support all data types, including numerical, categorical, and time series inputs (when treated as tabular data).
Some types (e.g., MNARType5 and MNARType6) are designed to operate on single columns only.
Numerical MNAR Types
Numerical Types Applies to continuous or ordinal variables. Includes self-masking, ranking-based, and quantile-based missingness strategies.
Categorical MNAR Types
Categorical Types Applies to discrete or binary variables. Supports class-conditional or value-dependent missingness patterns.