kNNSampler: Stochastic Imputations for Recovering Missing Value Distributions
This addresses the need for better uncertainty quantification in missing data imputation for data analysis applications, representing an incremental improvement over existing methods like kNNImputer.
The paper tackles the problem of missing-value imputation by proposing kNNSampler, a method that imputes missing values by randomly sampling from the observed responses of the k most similar units, enabling recovery of the conditional distribution of missing values and quantification of uncertainties, with experiments showing its effectiveness in distribution recovery.
We study a missing-value imputation method, termed kNNSampler, that imputes a given unit's missing response by randomly sampling from the observed responses of the $k$ most similar units to the given unit in terms of the observed covariates. This method can sample unknown missing values from their distributions, quantify the uncertainties of missing values, and be readily used for multiple imputation. Unlike popular kNNImputer, which estimates the conditional mean of a missing response given an observed covariate, kNNSampler is theoretically shown to estimate the conditional distribution of a missing response given an observed covariate. Experiments demonstrate its effectiveness in recovering the distribution of missing values. The code for kNNSampler is made publicly available (https://github.com/SAP/knn-sampler).