30 likes | 155 Vues
This project focuses on the automated design and management of online surveys, particularly in the social sciences, emphasizing quality control in survey responses. By implementing innovative metrics like the Survey Entropy Metric, we filter out random or careless respondents through the analysis of response patterns and consistency. Utilizing concepts like bootstrap resampling and Shannon entropy, we assess the significance of various responses in maintaining overall survey integrity. This research aims to improve the accuracy and reliability of survey data, ensuring more valid findings for academic researchers.
E N D
SurveyMan • Overall Project: automated design and management of online surveys using programming languages and systems concepts • Target Audience: Academic researchers, particularly in the Social Sciences • My Project: Quality Control – designing, implementing, and testing metrics to determine the quality of surveys and their responses Molly McMahon Advisor: Emery Berger Grad Mentor: Emma Tosch PLASMA Lab
Quality Control Purpose: • Filter out random/lazy respondents Existing strategies: • Infrequency • Answer patterns • Inconsistent responses • Responses times My current work: Survey Entropy Metric Concepts used: Bootstrap resampling, Shannon Entropy, Welch’s t test
Survey Entropy Metric • Given a survey and a list of responses, use bootstrapping to get resamples of the responses • Compute the entropy of each resampled list of responses, taking note of which original responses were not included in the resample • Compare the mean and standard deviation of all the resamples’ entropies to those of the resamples that did not include certain responses(Welch’s t-test) • If removing a certain response significantly lowers the survey entropy, response is an outlier