Bulletin of the Computational Statistics of Japan

Nonparametric techniques, such as kernel estimators and k-Nearest Neighbor (k-NN), are suitable for prediction techniques in large-scale data, since they directly hold the data structure for prediction. They can be robust techniques especially against changes of the data structure. However, the optimal bandwidth selection problem must be solved to enable high performance. Vieu (1991) proposed the optimal bandwidth selection technique for univariate data which is based on the minimization of a local cross-validation criterion. It is capable to be usefully applied for prediction problems. In this paper we expand the technique for multivariate data and show some numerical examples to clarify its validity.

View full abstract

Show abstractHide abstract

In many areas (e.g. sample survey, medical research, industrial experiment, etc.), researchers often encounter incomplete data for various reasons. Therefore, when designing a study, researchers must consider how to handle the incomplete data that might occur in their study. Recently the availability of software which provides statistical analysis for incomplete data has been increasing. In this paper we compare the functions and the characteristics of six software packages for statistical analyses for incomplete data and make a proposal on how to choose the software package depending on the purposes. With regard to the statistical methods, the main focus is on the EM algorithm (Dempster et al., 1977) and multiple imputation (Rubin, 1987). Finally, using randomly generated data, the output of some software packages are compared and also the differences in the results between statistical analysis methods for incomplete data are examined.

View full abstract

Download PDF (1316K)

A confidence interval on the difference of proportions is used in comparing effective rates of the medicine and so on. On the other hand, considering the data to a contingency table, we can use Fisher's exact test. However, the result of the test is not sometimes consistent with the confidence interval. Matsuda and Sano (2001) proposed a method of constructing the confidence interval which is consistent with the result of Fisher's test. In this paper, we show a new function on R that calculate the proposed confidence interval, and study quality of the confidence interval.

View full abstract

In this paper, we estimate Tokyo's land price for the last 23 years by universal kriging, a geostatistical method which provides the best linear unbiased predict for spatial data.Spatial correlation in disturbances is estimated through a semivariogram model and employed for spatial interpolation of official land prices in Tokyo. Our calculations show that spatial correlation in disturbances is crucial in estimating land prices and that kriging is one of the most appropriate approaches for this goal.

View full abstract

Register with J-STAGE for free!