ABSTRACT
Multiple Imputation (MI) was developed by Donald Rubin, and first described by him in a 1977 manuscript prepared for the United States Social Survey Administration. This is reproduced as an Appendix in his 1987 book Multiple Imputation for Nonresponse in Surveys (Rubin, 1987). The origins of MI clearly lie in the analysis of survey data, and the dominant (but not exclusive) paradigm for developing and justifying the approach, as set out in both the original manuscript and the subsequent book, is that of sample surveys. In the decades that followed, the use of MI has extended to many other application areas, most notably in biostatistics. As such it is now commonly framed in a likelihood-based frequentist paradigm.
