His early work on the theory of sufficient statistics was highly regarded at the time, and is still cited.
About 1950 he started developing a new, rather complicated theory of stochastic processes and behavior and, although he continued to do excellent work in other areas, including sufficient statistics and programming in operational research, his dominant research interest for the rest of his life was his process theory.
Most of his United States colleagues never really understood his approach to stochastic processes, but his work was highly regarded in Japan, and was published in several Japanese statistical journals.
His process theory was also highly regarded by colleagues at the University of New Mexico, where he spent several periods as visiting professor at the Institute of Mathematics.
His granddaughter, Elizabeth Alexandra Meghan Barankin, was born a year and a half after his death.