This text describes a method of constructing a statistical model when only incomplete data are available, and proposes specific estimation algorithms for solving various individual incomplete data problems. Coverage includes such topics as the use of the EM algorithm for applications in neural network technology; the processing of missing values commonly seen in multidimensional data; and estimation issues for robust and latent variable models. Watanabe teaches economics at Tokyo U., and Yamaguchi is with Rikkyo U. in Japan.
