CrowdMI: Multiple Imputation via Crowdsourcing

TL;DR

This paper introduces CrowdMI, a crowdsourcing-based method for imputing missing data by converting data into surveys, demonstrating comparable accuracy to traditional statistical models for both qualitative and quantitative data.

Contribution

The paper proposes a novel crowdsourcing approach for data imputation that replicates multiple imputation frameworks, offering an alternative to complex statistical models.

Findings

CrowdMI produces valid imputations for qualitative data.

CrowdMI achieves results comparable to statistical models.

The method is effective for both qualitative and quantitative missing data.

Abstract

Can humans impute missing data with similar proficiency as machines? This is the question we aim to answer in this paper. We present a novel idea of converting observations with missing data in to a survey questionnaire, which is presented to crowdworkers for completion. We replicate a multiple imputation framework by having multiple unique crowdworkers complete our questionnaire. Experimental results demonstrate that using our method, it is possible to generate valid imputations for qualitative and quantitative missing data, with results comparable to imputations generated by complex statistical models.