Estimation in nonprobability samples with Propensity Score Adjustment and Kernel Weighting
Conference
64th ISI World Statistics Congress - Ottawa, Canada
Format: IPS Abstract
Session: IPS 287 - Sample surveys in the era of Big Data and Machine Learning
Monday 17 July 10 a.m. - noon (Canada/Eastern)
Abstract
Selection bias in nonprobability samples has been widely studied by survey statisticians. Some of the most important methods are inverse probability weighting, mass imputation, doubly robust estimators and kernel smoothing. In this work we study new estimation techniques for nonprobability samples based on kernel weighting, which can be combined with machine learning methods. A case study involving a nonprobability survey is carried out to verify the performance of the proposed methodology.