All issues
Volume 36 (2007)
- Issue 2-3 Pages 71-
- Issue 1 Pages 1-
Volume 36, Issue 2-3
Displaying 1-5 of 5 articles from this issue
- |<
- <
- 1
- >
- >|
-
Manabu Kuroki2007 Volume 36 Issue 2-3 Pages 71-85
Published: December 30, 2007
Released on J-STAGE: June 12, 2009
JOURNAL FREE ACCESSIn this paper, we assume that cause-effect relationships between variables can be described as a directed acyclic graph and the corresponding linear structural equation model. Then, we consider the identification problem for total effects in the presence of unmeasured confounders. In order to solve the problem, the front door criterion (Pearl, 2000) and the conditional instrumental variable method (Brito and Pearl, 2002) are known as identifiability criteria for total effects. In this paper, we propose new graphical identifiability criteria for total effects based on those for factor models. The results of this paper are useful as the identifiability criteria for total effects and provide a new viewpoint to the identification problem of factor models.View full abstractDownload PDF (776K) -
Masayuki Ohkura, Toshinari Kamakura2007 Volume 36 Issue 2-3 Pages 87-98
Published: December 30, 2007
Released on J-STAGE: June 12, 2009
JOURNAL FREE ACCESSIn general, the maximum likelihood method is used to estimate regression parameters in logistic regression model. However, a maximum likelihood estimator does not exist in case of near separation. Furthermore, if the probability of occurring event is extremely small (or large) or the sample size is small compared with the number of regression parameters, a maximum likelihood estimator is not appropriate. In this situation, the exact logistic regression is useful. However, statistical softwares do not always support the exact logistic regression method. Firth (1993) suggested the method to remove bias of a maximum likelihood estimator, but unfortunately, this is not investigated well under near separation. The aim of this paper is to discuss the method to approximate a maximum likelihood estimator or an estimator using Firth's method to an estimator using the exact logistic regression. We also investigate the approximation method with relation to the data structure for improving regression estimates based on simulation study. It is shown that we can approximate a maximum likelihood estimator or an estimator using Firth's method to an estimator using the exact logistic regression. Furthermore, it is possible to improve accuracy of approximation for maximum likelihood estimator if we consider the data structure in the approximation.View full abstractDownload PDF (560K) -
Hiroki Motogaito, Tomoyuki Sugimoto, Masashi Goto2007 Volume 36 Issue 2-3 Pages 99-118
Published: December 30, 2007
Released on J-STAGE: June 12, 2009
JOURNAL FREE ACCESSIn regression problems, some of the most important goals are (i) to obtain a lower prediction error, and (ii) to interpret regression relationships. Friedman's Multivariate Adaptive Regression Splines (MARS) method which constructs basis functions with interaction effects is a very powerful data-driven technique in the viewpoint of (i), and the single tree-based structure built by MARS contributes to approach to (ii). Also, to address (i) and (ii), the better estimation and variable selection of the model are inevitable issues, and then shrinkage estimators contribute to resolve such issues. Recently, especially in the context of linear regression, Breiman's non-negative garrote (NNG) estimator and Tibshirani's Lasso estimator are shown to be a stable estimation and variable selection that often outperform theirs competitors. In this paper, we focus on Breiman's NNG as a foundation to incorporate the shrinkage estimators into the tree-based regression model and propose a new version of the MARS with the NNG (NNG-MARS). Then, we evaluated some performances of the NNG-MARS via an examination of a literature example and several small scale simulations. As a conclusion, the NNG-MARS, which holds the interpretable tree-based structure, has achieved much lower prediction error than the ordinary MARS.View full abstractDownload PDF (1075K) -
Takashi Daimon, Masashi Goto2007 Volume 36 Issue 2-3 Pages 119-137
Published: December 30, 2007
Released on J-STAGE: June 12, 2009
JOURNAL FREE ACCESSIn clinical trials, an interim monitoring is conducted so as to consider the early termination/continuation or design modification of a trial, from the ethical, administrative, and financial viewpoints. Statistical approaches to the interim monitoring can be, in general, classified into the frequentist and Bayesian paradigms. In this article we focus on the interim monitoring in the latter. In the Bayesian interim monitoring, one must construct the model that provides the posterior information by combining the data information with the prior, like the general Bayesian approaches, and judge the pros and cons of early termination/continuation of a trial, based on a index derived from the model. Then it is required to check whether the model and derived index are reliable or not. In this paper, a predictive checking approach is proposed, and its performance are evaluated in several examples. The results showed that the predictive checking approach allowed us to evaluate the consistency in information between the prior, data, and posterior and check the reliability of the model and indices.View full abstractDownload PDF (1079K) -
Yosihito Maruyama2007 Volume 36 Issue 2-3 Pages 139-145
Published: December 30, 2007
Released on J-STAGE: June 12, 2009
JOURNAL FREE ACCESSThis paper is concerned with comparison of two types of multivariate sample skewness and kurtosis, which are Mardia-type and Srivastava-type, respectively. Asymptotic expansions under non-normality are obtained for the sample measures of Srivastava-type in order to consider their asymptotic behaviors.View full abstractDownload PDF (265K)
- |<
- <
- 1
- >
- >|