Abstract:
|
Higher perception of respondent’s subjective burden can potentially induce uncertain responses to a questionnaire or refusals to continue participation in a survey which in turn can introduce bias and downgrade the quality of overall survey data. Interventions may be required during data collection to maintain data quality to reduce respondent’s subjective burden along with its potential bias. In this study, we use the Consumer Expenditure Surveys (CE) Interview data collected between April 2017 and March 2018to explore how nonparametric recursive partitioning random forest models can be used to predict a subjective burden outcome from inputs of objective burden measures, household demographics and other explanatory variables under a complex survey design.
|