Online Program Home
My Program

Abstract Details

Activity Number: 237 - SPEED: Missing Survey Data: Analysis, Imputation, Design, and Prevention
Type: Contributed
Date/Time: Monday, July 30, 2018 : 2:00 PM to 3:50 PM
Sponsor: Survey Research Methods Section
Abstract #330105
Title: Imputation of Small Number of New Questions in the Large Survey
Author(s): Di Xiong* and Yan Wang and Honghu Liu
Companies: UCLA SPH and Field School of Public Health, UCLA and UCLA
Keywords: Bootstrap; Multiple Imputation; CART; Sensitivity; Specificity

It is common to test very few additional survey questions towards the end of the study. Usually, these questions were answered by a small number of pilot subjects, but have a large number of questions overlapped with the main survey answered by the study population. In this paper, we aim to investigate the methods to impute the response of the study population for these newly added survey questions. Bootstrap methods and Classification and Regression Tree (CART) are applied to impute the missing block of the response matrix for these very few questions. The relationship between the imputed responses and the outcome variables (the gold standard that the survey questions intended to measure) can be used to compare the imputation results with pilot results, e.g. sensitivity and specificity. Finally, the method is applied to a pediatric oral health survey research to evaluate the imputed responses with visual dental exam results. We compare the sensitivity and specificity of the imputed set with the small complete data. The imputation results can inform whether the test questions may have a higher predicted power for the outcome variable when applied to a larger population.

Authors who are presenting talks have a * after their name.

Back to the full JSM 2018 program