• Journal of the Korean Statistical Society
• /
• v.31 no.1
• /
• pp.1-16
• /
• 2002

Since changepoint identification is important in many data analysis problem, we wish to make inference about the locations of one or more changepoints of the sequence. We consider the Bayesian nonparameteric inference for multiple changepoint problem using a Bayesian segmentation procedure proposed by Yang and Kuo (2000). A mixture of products of Dirichlet process is used as a prior distribution. To decide whether there exists a single change or not, our approach depends on nonparametric Bayesian Schwartz information criterion at each step. We discuss how to choose the precision parameter (total mass parameter) in nonparametric setting and show that the discreteness of the Dirichlet process prior can ha17e a large effect on the nonparametric Bayesian Schwartz information criterion and leads to conclusions that are very different results from reasonable parametric model. One example is proposed to show this effect.

• Journal of Korean Society of Coastal and Ocean Engineers
• /
• v.19 no.2
• /
• pp.170-178
• /
• 2007

Most of parabolic approximation models employ a relatively limited open boundary condition in which there is no depth variation in the longshore direction outside of the computation domain so that Snell's law may be presumed to hold. Existing Kirby's condition belongs to this category and in the paper both modified Kirby's method and Dirichlet boundary condition are presented in detail and numerical results of three methods were shown. Judging from computation to wave propagations over a circular shoal in a constant depth, the method based on present Dirichlet boundary condition with fictitious numerical adjusting regions in both sides of the computation domain gives the least distorted amplitude ratio distribution.

• Journal of Distribution Research
• /
• v.4 no.1
• /
• pp.17-29
• /
• 1999

We investigated the consumer behavior of store choice using consumer panel data. The NBD-Dirichlet model known to be predictive of the consumer's brand choice was also found to be well fitted for the store choice behavior. Understanding the regularity in the store choice will provide both manufacturers and sistributors with the necessary guidelines for their competitive strategies.

• Journal of the Korean Statistical Society
• /
• v.30 no.3
• /
• pp.481-498
• /
• 2001

Empirical Bayes procedure of nonparametric estiamtion of cumulative hazard rates based on censored data is considered using the beta process priors of Hjort(1990). Beta process priors with unknown parameters are used for cumulative hazard rates. Empirical Bayes estimators are suggested and asymptotic optimality is proved. Our result generalizes that of Susarla and Van Ryzin(1978) in the sensor that (i) the cumulative hazard rate induced by a Dirichlet process is a beta process, (ii) our empirical Bayes estimator does not depend on the censoring distribution while that of Susarla and Van Ryzin(1978) does, (iii) a class of estimators of the hyperprameters is suggested in the prior distribution which is assumed known in advance in Susarla and Van Ryzin(1978). This extension makes the proposed empirical Bayes procedure more applicable to real dta sets.

• Communications for Statistical Applications and Methods
• /
• v.14 no.1
• /
• pp.121-131
• /
• 2007

In this paper, we examine the problem of estimating the sensitive characteristics and behaviors in a multinomial randomized response model using Bayesian approach. We derived a posterior distribution for parameter of interest for multinomial randomized response model. Based on the posterior distribution, we also calculated a credible intervals and mean squared error (MSE). We finally compare the maximum likelihood estimator and the Bayes estimator in terms of MSE.

• The Pure and Applied Mathematics
• /
• v.5 no.2
• /
• pp.123-132
• /
• 1998

Let F and G denote the distribution functions of the failure times and the censoring variables in a random censorship model. Susarla and Van Ryzin(1978) verified consistency of $F_{\alpha}$, he NPBE of F with respect to the Dirichlet process prior D($\alpha$), in which they assumed F and G are continuous. Assuming that A, the cumulative hazard function, is distributed according to a beta process with parameters c, $\alpha$, Hjort(1990) obtained the Bayes estimator $A_{c,\alpha}$ of A under a squared error loss function. By the theory of product-integral developed by Gill and Johansen(1990), the Bayes estimator $F_{c,\alpha}$ is recovered from $A_{c,\alpha}$. Continuity assumption on F and G is removed in our proof of the consistency of $A_{c,\alpha}$ and $F_{c,\alpha}$. Our result extends Susarla and Van Ryzin(1978) since a particular transform of a beta process is a Dirichlet process and the class of beta processes forms a much larger class than the class of Dirichlet processes.

• The Korean Journal of Applied Statistics
• /
• v.29 no.4
• /
• pp.627-641
• /
• 2016

In this paper, we consider a seemingly unrelated regression (SUR) model and propose a nonparametric Bayesian approach to SUR with a Dirichlet process mixture of normals for modeling an unknown error distribution. Posterior distributions are derived based on the proposed model, and the posterior inference is performed via Markov chain Monte Carlo methods based on the collapsed Gibbs sampler of a Dirichlet process mixture model. We present a simulation study to assess the performance of the model. We also apply the model to precipitation data over South Korea.

• Smart Media Journal
• /
• v.11 no.3
• /
• pp.9-17
• /
• 2022

Algorithmic trading, which uses algorithms to trade financial products, has a problem in that the results are not stable due to many factors in the market. To alleviate this problem, ensemble techniques that combine trading algorithms have been proposed. However, there are several problems with this ensemble method. First, the trading algorithm may not be selected so as to satisfy the minimum performance requirement (more than random) of the algorithm included in the ensemble, which is a necessary requirement of the ensemble. Second, there is no guarantee that an ensemble model that performed well in the past will perform well in the future. In order to solve these problems, a method for selecting trading algorithms included in the ensemble model is proposed as follows. Based on past data, we measure the contribution of the trading algorithms included in the ensemble models with high performance. However, for contributions based only on this historical data, since there are not enough past data and the uncertainty of the past data is not reflected, the contribution distribution is approximated using the Dirichlet distribution, and the contribution values are sampled from the contribution distribution to reflect the uncertainty. Based on the contribution distribution of the trading algorithm obtained from the past data, the Transformer is trained to predict the future contribution. Trading algorithms with high predicted future contribution are selected and included in the ensemble model. Through experiments, it was proved that the proposed ensemble method showed superior performance compared to the existing ensemble methods.

• Communications for Statistical Applications and Methods
• /
• v.21 no.2
• /
• pp.135-145
• /
• 2014

The present study surveys Bayesian modeling structure for inferences about transition probabilities of Markov chain. The motivation of the study came from the data that shows transitional behaviors of emotionally disturbed children undergoing residential treatment program. Dirichlet distribution was used as prior for the multinomial distribution. The analysis with real data was implemented in WinBUGS programming environment. The performance of the model was compared to that of alternative approaches.

• ETRI Journal
• /
• v.43 no.1
• /
• pp.74-81
• /
• 2021

The mixture model is a very powerful and flexible tool in clustering analysis. Based on the Dirichlet process and parsimonious Gaussian distribution, we propose a new nonparametric mixture framework for solving challenging clustering problems. Meanwhile, the inference of the model depends on the efficient online variational Bayesian approach, which enhances the information exchange between the whole and the part to a certain extent and applies to scalable datasets. The experiments on the scene database indicate that the novel clustering framework, when combined with a convolutional neural network for feature extraction, has meaningful advantages over other models.