My chancy life as a Statistician
Section 4. Life in Canada: 1959-2000
I found that the Canadian universities suited my family circumstances in India at that time and decided to migrate to Canada in 1969 directly from Calcutta. Hartley was very unhappy with my decision, but we continued our collaboration for several years. I worked four years at the University of Manitoba before joining Carleton University, Ottawa in 1969. I have also worked at Statistics Canada for the past 40 years or so as a consultant, and this practical exposure was extremely useful in my later research work. I have collaborated with many statisticians over the past 25 years, thanks to my Canadian NSERC research grant that encourages collaborative work. I supervised many outstanding Ph.D. students in Canada. My first Ph.D. student in Canada, David Bellhouse (co-supervised with Jim Kalbfleish at the University of Waterloo), wrote his thesis on optimal estimation in finite population sampling. He had a distinguished career at the University of Western Ontario and retired recently. Bellhouse is also a leading expert in the history of Statistics. Dan Krewski was my first Ph.D. student at Carleton University. He developed asymptotic theory for stratified multistage sampling designs (Krewski and Rao, 1981) which provided theoretical justification for replication methods, such as the jackknife and balanced repeated replication, widely used for the analysis of complex survey data (see Shao and Tu, 1995, Chapter 6). Krewski is currently a distinguished professor of biostatistics and population health at the University of Ottawa and he is a leading authority on risk assessment. Both Bellhouse and Krewski are ASA Fellows. Several of my Masters and Ph.D. students established successful careers at Statistics Canada and elsewhere.
In 1977, I was looking for a suitable place to spend my sabbatical leave. By chance, I bumped into Fred Smith of the University of Southampton at a survey sampling conference held at the University of North Carolina. He mentioned that he has applied for a research project on the analysis of complex survey data and if successful I could spend my sabbatical leave at his university working on the project. His research project was approved and I joined the project team (Fred Smith, Tim Holt, Gad Nathan and Alastair Scott) in April 1978 for 4 months. I also had a chance to interact with Graham Kalton who was also at the University of Southampton. I might mention that Smith, Holt and Kalton developed a strong program in survey sampling research at the University of Southampton. In later years, Chris Skinner, Ray Chambers and Danny Pfeffermann contributed greatly and made it into a leading center for survey research.
During my sabbatical leave, Alastair Scott and I worked on methods for the analysis of categorical survey data and published several papers subsequently. In Rao and Scott (1981, 1984), we developed simple corrections to standard chi-squared tests for testing independence in a two-way table of weighted counts that account for the survey design features. It was nice to see the 1981 paper with Scott included among the 19 landmark papers in survey sampling published over the period 1930-90. Scott visited me regularly for several years to continue our work on analysis of survey data and other topics until his health did not permit him to travel alone. He was suffering from brain cancer but hoped to attend the China conference in May 2017. I was deeply saddened by the news of his death on the first day of the conference. I would like to dedicate this joint special issue of ISR and Survey Methodology to the memory of my dear friend and collaborator, Alastair Scott.
I collaborated with several excellent researchers after my return from sabbatical leave. Jeff Wu and I developed valid bootstrap variance estimators for stratified multistage sampling and other designs (Rao and Wu, 1988) and we introduced the concept of bootstrap weights (Rao, Wu and Yue, 1992). Currently, bootstrap weights are used at Statistics Canada for variance estimation in several large-scale surveys. Other major collaborations include the following: (1) multiple frame surveys with Chris Skinner, Sharon Lohr and Changbao Wu, (2) empirical likelihood intervals for survey data with Changbao Wu, Jiahua Chen, Yves Berger and M. Salehi, (3) analysis of survey data with Alastair Scott, Chris Skinner, Roland Thomas, Mike Hidiroglou, Wesley Yung and Jun Shao, (4) imputation for missing data with Jun Shao, Randy Sitter, Jae Kim, Qihua Wang, Jiahua Chen and Y.S. Qin. Randy Sitter and Jun Shao were my colleagues during the period 1990-95, and our statistics group was rated among the top 15 in the world for research productivity. Other collaborators include Arun Nigam, Jurgen Kleffe, K. Vijayan, Avi Singh, Gordon Brackstone, Poduri Rao and P.A.V.B. Swamy.
Around 1985, I got interested in small area estimation after organizing an international symposium on small area statistics in 1985 jointly with Statistics Canada. Invited papers presented at the symposium are published in a Wiley book (Platek, Rao, Särndal and Singh, 1987). Demand for reliable small area statistics has steadily grown in the past 25 years which in turn led to many theoretical and practical contributions. I supervised several Ph.D. students on this topic, including N.G.N. Prasad, Diane Stukel, Ming Yu and Yong You. Prasad developed accurate mean squared error estimators of model-based small area estimators (Prasad and Rao, 1990) and this work is widely cited. Yong You received the Pierre Robillard award of the Statistical Society of Canada for the best Ph.D. thesis in the year he graduated.
- Date modified: