CrowdMI: Multiple Imputation via Crowdsourcing
This addresses the data imputation problem for researchers and practitioners by offering a human-based alternative to machine methods, though it appears incremental as it adapts existing multiple imputation frameworks to crowdsourcing.
The paper tackled the problem of imputing missing data by proposing a crowdsourcing approach where crowdworkers complete surveys based on observations with missing data, and found that this method generates valid imputations for both qualitative and quantitative data, with results comparable to those from complex statistical models.
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.