Dealing with missing data in online ‘opt-in’ non-probability panels

Project Details

Lead Supervisor: Natalie Shlomo, University of Manchester
Other Supervisors: Alexandru Cernat, Mark Elliot
Contact Email:
Partners: University of Manchester
External Partners: YouGov

Start Date: October 2017

Online ‘opt-in’ non-probability panels are increasingly being used by polling and marketing agencies due to their low costs and timeliness.  Unlike traditional probability surveys they sample from a dataset of ‘opt-in’ panel members (which can be larger than a million for some companies) in a relatively high frequency fashion (e.g., daily). Despite their popularity they also present some challenges which mainly deal with how to obtain correct and unbiased inference. Moreover, nonresponse and missing data can be relatively high. Traditional methods for compensating for missing data such as drawing from the empirical distribution of the data may not be applicable due to the non-random nature of the data. The project will explore  new methods for compensating for missing data such as model based or computer learning algorithms.

Reference number MN02

Deadline for applications – 30th April 2017

Apply online here

Tags: , , , ,