山下 遥

This article proposes a method for estimating principal points for a multivariate binary distribution, assuming a log-linear model for the distribution. Through numerical simulation studies, the proposed parametric estimation method using a log-linear model is compared with a nonparametric estimation method.

Principal points for binary distributions are able to be defined based on Flury's principal points (1990). However, finding principal points for binary distributions is hard in a straightforward manner. In this article, a method for approximating principal points for binary distributions is proposed by formulating it as an uncapacitated location problem. Moreover, it is shown that the problem of finding principal points can be solved with the aid of submodular functions. It leads to a solution whose value is at least (1 - 1/e) times the optimal value.

Recently, a parametric estimation method for principal points for a multivariate binary distribution using a log-linear model has been proposed, and Akaike information criterion (AIC) has been applied to model selection for log-linear model. This paper compares three model selection methods based on AIC, Bayesian information criterion (BIC), and the likelihood ratio test (LRT) for estimating principal points for a multivariate binary distribution. The performances of the model selection methods are shown through numerical simulation studies

The analysis of binary (0 or 1) data requires an analysis method whose objects are realizations. Yamashita and Suzuki (to appear) proposed principal points for binary distributions based on the concept of principal points, defined by Flury (1990). Ideally, when we search for the binary principal points, all combinations of the k-principal points should be considered; however, this problem cannot be solved in a straightforward manner because the number of combinations increases exponentially when the number of the variables increases. In this paper, we propose three heuristic methods for approximating principal points for binary distributions. The results indicate that our method enables us to find approximated principal points and summarize a binary distribution using the points.