Professor Kim Jae-kwang at Iowa State University proposed a fractional imputation method as a way of handling missing data. This is my interview with him.
Q> You are one of the recipients of the Gertrude M Cox award. Can you explain more about this award?
I received the award about 5 years ago. The award annually recognizes a statistician in their early to mid-career (fewer than 15 years after getting PhD degree) who has made significant contributions to one or more of the applied statistics areas in which Gertrude Cox (1900–1978) worked: survey methodology, experimental design, biostatistics, and statistical computing. I was the first Korean to receive this award. Recently, I was nominated as a fellow of the Institute of Mathematical Statistics. There are three Korean fellows of IMS so far. Professors Byung U Park and Yongdai Kim are the other two Korean statisticians who were elected to be fellows of IMS.
Q>Can you briefly introduce your career path before you joined Iowa State University (ISU)?
I got my Ph.D. degree at ISU in 2000 and worked in a company in the US for two years. After that, I worked as an Assistant Professor at the Hankuk University of Foreign Studies (for two years) and at Yonsei University (for 4.5 years).
Q> What are the difficulties for getting a faculty job at ISU?
I did not have any teaching experience in English-only courses. So, I was unsure of my teaching abilities in English. That made me hesitate to accept the job offer from ISU.
Q> Can you briefly introduce your research contribution?
I originally majored in survey sampling in statistics. In data collection, the nonresponse rate gets higher and the need for handling nonresponse in a scientific way is increasing. One of the popular methods for handling nonresponse is multiple imputations, proposed by Professor Donald Rubin at Harvard University. I discovered some problems in using multiple imputations for general-purpose estimations, and then theoretically showed why it could be problematic. Also, with my collaborators, I proposed fractional imputation as an alternative tool for handling missing data.
The fractional imputation uses a frequentist framework, rather than the Bayesian framework as in multiple imputations. At first, Professor Rubin and his disciples refused to accept my ideas. However, fractional imputation is now used in many applications. Also, I worked on some theoretical investigations of handling missing data with non-ignorable nonresponse.
I also worked on several real data consulting projects sponsored by government agencies in the US. For example, I worked on developing a fractional imputation method for handling missing income items for the US Census Bureau. I also worked with the National Agricultural Statistical Services to develop a small area estimation method for predicting crop acreages incorporating the satellite image classification data and the survey data in the June Area Survey.
Q> How does your job experience differ from that of universities in Korea?
When I worked at Yonsei University, I felt that I would settle for a comfortable life as a professor in a top university in Korea. However, in the US, I did not have that same feeling of comfort, and I could not settle down.
Q> What is your opinion on the viewpoint of statistics as a branch of mathematics?
I do not have a real strong opinion about the classification of majors and subject areas. We do not have to restrict our interest to the area that we are comfortable with. In fact, many practical problems require multi-disciplinary research skills. So, collaboration and willingness to communicate with experts in other areas are more important attitudes than ever before.
Q> Can you explain the importance of statistics?
The importance of statistics is well accepted and will be fully appreciated by whoever wishes to discover knowledge from data. Any domain areas using data will need statistical techniques.
Q> Do you have any advice on our next generation who wants to find a job in the US?
I am not sure how many would be interested in moving out to the US under the Trump presidency (^^), but it is a young man’s privilege to challenge himself before getting settled. Also, as a researcher, working one’s way up in a top tier research school in the US is worthy of the effort.
Q> What is your next academic plan?
I am writing the second edition of the book “Statistical Methods for Handling Incomplete Data”. Once the book is written, I would like to study a somewhat different research area, such as machine learning or data integration.
Q> Can you share your most memorable experience?
Looking back, I am satisfied overall. Having family trips during the summer to National Parks or sightseeing places in Europe is probably the most pleasant time. My kids have grown up now. They no longer want to go on trips with me!
