Simple Surveys: Response Retrieval Inspired by Recommendation Systems

Sengupta, Nandana; Srebro, Nati; Evans, James E.

doi:10.48550/arxiv.1901.09659

Cited by 1 publication

(1 citation statement)

References 0 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Multiple authors have suggested applying matrix completion to survey imputation [Candès and Recht (2009); Davenport et al (2014); Klopp et al (2015); Josse et al (2016)], but published applications to real-world surveys are rare. One exception is concurrent work by Sengupta et al (2018) that examines the predictive ability of matrix completion on survey responses collected by different elicitation strategies.…”

Section: Introduction 1reducing Response Burden In Surveysmentioning

confidence: 99%

Active Matrix Factorization for Surveys

Zhang¹,

Taylor²,

Cobb³

et al. 2019

Preprint

View full text Add to dashboard Cite

Amid historically low response rates, survey researchers seek ways to reduce respondent burden while measuring desired concepts with precision. We propose to ask fewer questions of respondents and impute missing responses via probabilistic matrix factorization. A variance-minimizing active learning criterion chooses the most informative questions per respondent. In simulations of our matrix sampling procedure on real-world surveys, as well as a Facebook survey experiment, we find active question selection achieves efficiency gains over baselines. The reduction in imputation error is heterogeneous across questions, and depends on the latent concepts they capture. The imputation procedure can benefit from incorporating respondent side information, modeling responses as ordered logit rather than Gaussian, and accounting for order effects. With our method, survey researchers obtain principled suggestions of questions to retain and, if desired, can automate the design of shorter instruments.

show abstract