Collaborative Research Team Project: 2015-2018
Complex surveys play an important role in providing information for science and society. How to best handle missing data in surveys has been one of the focal points of research in the past three decades, with the ultimate prize of achieving reliable and efficient use of information from complex surveys. Imputation is an increasingly used technique for missing data. However, it is notoriously difficult to impute multivariate data having an arbitrary missing pattern if the interest lies in preserving the covariance structure in the imputed data, and addressing this problem will be a major part of the planned research.
This project will carry out research and training in survey data analysis necessitated by the advent of increasingly extensive data sets, and challenges in data collection resulting from non-response and missing values. The team will collaborate with researchers at Statistics Canada and Westat (Rockville, USA) on problems in large-scale, complex and high-dimensional survey data with missing values. They plan to extend the techniques of fractional imputation and doubly robust methods to dealing with missing values in high dimensional data. They will also explore the relatively new area of developing inference from incomplete functional survey data, with application to large functional data sets, such as electricity consumption data.
The team leader is David Haziza of Université de Montréal, with collaborators Jean-François Beaumont of Statistics Canada, Michael Brick of Westat, Hervé Cardot and Camelia Goga of Université de Bourgogne, Jiahua Chen of University of British Columbia, Jae-Kwang Kim of Iowa State University, Wilson Lu of Acadia University, and Changbao Wu of the University of Waterloo.