• The Korean Journal of Applied Statistics
• /
• v.15 no.1
• /
• pp.139-151
• /
• 2002

This study is concerned with model selection and diagnostics for nonlinear regression model through Bayes factor. In this paper, we use informative prior and simulate observations from the posterior distribution via Markov chain Monte Carlo. We propose the Laplace approximation method and apply the Laplace-Metropolis estimator to solve the computational difficulty of Bayes factor.

• The Transactions of the Korea Information Processing Society
• /
• v.6 no.8
• /
• pp.2060-2071
• /
• 1999

The complicate software failure system is defined to the superposition of the points of failure from several component point process. Because the likelihood function is difficulty in computing, we consider Gibbs sampler using iteration sampling based method. For each observed failure epoch, we applied to latent variables that indicates with component of the superposition mode. For model selection, we explored the posterior Bayesian criterion and the sum of relative errors for the comparison simple pattern with superposition model. A numerical example with NHPP simulated data set applies the thinning method proposed by Lewis and Shedler[25] is given, we consider Goel-Okumoto model and Weibull model with GOS, inference of parameter is studied. Using the posterior Bayesian criterion and the sum of relative errors, as we would expect, the superposition model is best on model under diffuse priors.

• The Korean Journal of Applied Statistics
• /
• v.18 no.1
• /
• pp.1-13
• /
• 2005

Latent Class model has been considered recently by many researchers and practitioners as a tool for identifying heterogeneous segments or groups in a population, and grouping objects into the segments. In this paper we consider data on prostate cancer patients from Korean National Cancer Institute and propose a method for grouping prostate cancer patients by using latent class Poisson model. A Bayesian approach equipped with a Markov chain Monte Carlo method is used to overcome the limit of classical likelihood approaches. Advantages of the proposed Bayesian method are easy estimation of parameters with their standard errors, segmentation of objects into groups, and provision of uncertainty measures for the segmentation. In addition, we provide a method to determine an appropriate number of segments for the given data so that the method automatically chooses the number of segments and partitions objects into heterogeneous segments.

• The Korean Journal of Applied Statistics
• /
• v.35 no.2
• /
• pp.285-298
• /
• 2022

Continuous shrinkage priors, as well as spike and slab priors, have been widely employed for Bayesian inference about sparse regression coefficient vectors or covariance matrices. Continuous shrinkage priors provide computational advantages over spike and slab priors since their model space is substantially smaller. This is especially true in high-dimensional settings. However, variable selection based on continuous shrinkage priors is not straightforward because they do not give exactly zero values. Although few variable selection approaches based on continuous shrinkage priors have been proposed, no substantial comparative investigations of their performance have been conducted. In this paper, We compare two variable selection methods: a credible interval method and the sequential 2-means algorithm (Li and Pati, 2017). Various simulation scenarios are used to demonstrate the practical performances of the methods. We conclude the paper by presenting some observations and conjectures based on the simulation findings.

• Journal of the Korean Statistical Society
• /
• v.30 no.2
• /
• pp.179-192
• /
• 2001

When a statistical model has a hierarchical structure such as multilayer perceptrons in neural networks or Gaussian mixture density representation, the model includes distribution with unidentifiable parameters when the structure becomes redundant. Since the exact structure is unknown, we need to carry out statistical estimation or learning of parameters in such a model. From the geometrical point of view, distributions specified by unidentifiable parameters become a singular point in the parameter space. The problem has been remarked in many statistical models, and strange behaviors of the likelihood ratio statistics, when the null hypothesis is at a singular point, have been analyzed so far. The present paper studies asymptotic behaviors of the maximum likelihood estimator and the Bayesian predictive estimator, by using a simple cone model, and show that they are completely different from regular statistical models where the Cramer-Rao paradigm holds. At singularities, the Fisher information metric degenerates, implying that the cramer-Rao paradigm does no more hold, and that he classical model selection theory such as AIC and MDL cannot be applied. This paper is a first step to establish a new theory for analyzing the accuracy of estimation or learning at around singularities.

• Journal of KIISE:Software and Applications
• /
• v.30 no.9
• /
• pp.867-875
• /
• 2003

This paper concerns continuous density HMM topology optimization. There have been several researches for HMM topology optimization. BIC (Bayesian Information Criterion) is one of the well known optimization criteria, which assumes statistically well behaved homogeneous model parameters. HMMs, however, are composed of several different kind of parameters to accommodate complex topology, thus BIC's assumption does not hold true for HMMs. Even though BIC reduced the total number of parameters of HMMs, it could not improve the recognition rates. In this paper, we proposed two new model selection criteria, HBIC (HMM-oriented BIC) and BIC_Anti. The former is proposed to improve BIC by estimating model priors separately. The latter is to combine BIC and anti-likelihood to accelerate discrimination power of HMMs. We performed some comparative research on couple of model selection criteria for online handwriting data recognition. We got better recognition results with fewer number of parameters.

• Animal Bioscience
• /
• v.34 no.2
• /
• pp.185-191
• /
• 2021

Objective: The objective of this study was to estimate the genetic parameters for worm resistance (WR) and associated characteristics, using the linear-threshold animal model via Bayesian inference in single- and multiple-trait analyses. Methods: Data were collected from a herd of Santa Inês breed sheep. All information was collected with animals submitted to natural contamination conditions. All data (number of eggs per gram of feces [FEC], Famacha score [FS], body condition score [BCS], and hematocrit [HCT]) were collected on the same day. The animals were weighed individually on the day after collection (after 12-h fasting). The WR trait was defined by the multivariate cluster analysis, using the FEC, HCT, BCS, and FS of material collected from naturally infected sheep of the Santa Inês breed. The variance components and genetic parameters for the WR, FEC, HCT, BCS, and FS traits were estimated using the Bayesian inference under the linear and threshold animal model. Results: A low magnitude was obtained for repeatability of worm-related traits. The mean values estimated for heritability were of low-to-high (0.05 to 0.88) magnitude. The FEC, HCT, BCS, FS, and body weight traits showed higher heritability (although low magnitude) in the multiple-trait model due to increased information about traits. All WR characters showed a significant genetic correlation, and heritability estimates ranged from low (0.44; single-trait model) to high (0.88; multiple-trait model). Conclusion: Therefore, we suggest that FS be included as a criterion of ovine genetic selection for endoparasite resistance using the trait defined by multivariate cluster analysis, as it will provide greater genetic gains when compared to any single trait. In addition, its measurement is easy and inexpensive, exhibiting greater heritability and repeatability and a high genetic correlation with the trait of resistance to worms.

• Journal of the Korean Statistical Society
• /
• v.35 no.4
• /
• pp.419-439
• /
• 2006

The article develops a new class of distributions by introducing a nonnegative perturbing function to $t_\nu$ distribution having location and scale parameters. The class is obtained by using transformations and conditioning. The class strictly includes $t_\nu$ and $skew-t_\nu$ distributions. It provides yet other models useful for selection modeling and robustness analysis. Analytic forms of the densities are obtained and distributional properties are studied. These developments are followed by an easy method for estimating the distribution by using Markov chain Monte Carlo. It is shown that the method is straightforward to specify distribution ally and to implement computationally, with output readily adopted for constructing required criterion. The method is illustrated by using a simulation study.

• The Korean Journal of Applied Statistics
• /
• v.14 no.2
• /
• pp.449-462
• /
• 2001

베이지안 다중 검정방법(multiple hypothesis test)은 여러 통계모형에서 성공적인 결과를 주는 것으로 알려져있다. 일반적으로, 베이지안 가설검정은 고려중인 모형에 대한 사후확률을 계산하여 가장 높은 확률은 갖는 모형을 선택하기 때문에 귀무가설의 기각여부에만 관심을 가지는 고전적인 분산분석 검정과는 달리 좀 더 구체적인 모형을 선택할 수 있는 장점이 있다. 이 논문에서는 독립이면서 로그정규분포를 따르는 K($\geq$3)개 모집단의 모수에 대한 가설 검정방법으로 O’Hagan(1995)이 제안한 부분 베이즈 요인을 이용한 베이지안 방법을 제안한다. 이 때 모수에 대한 사전분포로는 무정보적 사전분포를 사용한다. 제안한 검정 방법의 유용성을 알아보기 위하여 실제 자료의 분석과 모의 실험을 이용하여 고전적인 검정방법과 그 결과를 비교한다.

• Journal of the Korean Statistical Society
• /
• v.29 no.4
• /
• pp.443-454
• /
• 2000

In Bayesian model selection or testing problems of different dimensions, the conventional Bayes factors with improper noninformative priors are not well defined. The intrinsic Bayes factor and the fractional Bayes factor are used to overcome such problems by using a data-splitting idea and fraction, respectively. This article addresses a Bayesian testing for the comparison of two normal means with unknown variance. We derive proper intrinsic priors, whose Bayes factors are asymptotically equivalent to the corresponding fractional Bayes factor. We demonstrate our results with two examples.