kNNSampler: Stochastic Imputations for Recovering Missing Value Distributions

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Pashmchi, Parastoo, Benoit, Jérôme, Kanagawa, Motonobu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866909939436355584
author Pashmchi, Parastoo
Benoit, Jérôme
Kanagawa, Motonobu
author_facet Pashmchi, Parastoo
Benoit, Jérôme
Kanagawa, Motonobu
contents 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 illustrate the performance of kNNSampler. The code for kNNSampler is made publicly available (https://github.com/SAP/knn-sampler).
format Preprint
id arxiv_https___arxiv_org_abs_2509_08366
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle kNNSampler: Stochastic Imputations for Recovering Missing Value Distributions
Pashmchi, Parastoo
Benoit, Jérôme
Kanagawa, Motonobu
Machine Learning
Statistics Theory
Methodology
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 illustrate the performance of kNNSampler. The code for kNNSampler is made publicly available (https://github.com/SAP/knn-sampler).
title kNNSampler: Stochastic Imputations for Recovering Missing Value Distributions
topic Machine Learning
Statistics Theory
Methodology
url https://arxiv.org/abs/2509.08366