Proceedings of the symposium of Japanese Society of Computational Statistics

Pathway-based approach for biomarker selection(Session 5C(IASC-ARS))

Ya-Hsuan Chang, Chung-Ming Chen, Hsuan-Yu Chen

Article type: Article
Pages 173-176
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_173

CONFERENCE PROCEEDINGS FREE ACCESS

Download PDF (265K)
Comparing Correlated ROC Curves By A Two Stage Test(Session 5C(IASC-ARS))

Wenbao Yu, Eunsik Park

Article type: Article
Pages 177-180
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_177

CONFERENCE PROCEEDINGS FREE ACCESS

Show abstractHide abstract

The area under the ROC curve (AUC) is the most popular index when comparing two ROC curves and the significant test of two areas have been well developed. However, this index is invalid when the ROC curves cross while the underline AUCs are similar. In this paper, a simple two stage nonparametric significant test is proposed for comparing ROC curves in paired designs. The test is composed by an non-parametric AUC test and a non-parametric modified AUC (mAUC) test. The new procedure is shown by simulations having higher power than two conventional ROC-type tests, especially when the two ROC curves cross and have close areas. Moreover, the proposed test also implicitly tests the equality of two binormal ROC curves, which is the most popular parametric ROC curve model, at every operating point.

View full abstract

Download PDF (374K)
PARTITION METHOD IN COPULA BASED HIERARCHICAL BAYESIAN SPATIO-TEMPORAL MODEL(Session 5C(IASC-ARS))

Amran, N Iriawan, Subiono, Irhamah

Article type: Article
Pages 181-184
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_181

CONFERENCE PROCEEDINGS FREE ACCESS

Show abstractHide abstract

Copula Based Hierarchical Bayesian Spatio-Temporal Model is a Copula based Spatio-Temporal model using Hierarchical Bayesian method in parameter estimation. Hierarchical Bayesian method estimates parameters of model in three sequential steps, i.e., Data model, Process model, and Priors. In the data model step, marginal density of observation is joined by one copula function. In this paper, we propose partition method to make homogeneity clusters in order to apply Copula Based Hierarchical Bayesian Spatio-Temporal Model at non-homogeneity extreme rainfalls observation. Using rainfalls data at twenty seventh stations rain gauge, the proposed method showed good performance. There are three different characteristics of extreme rainfalls were found by this method. The results can be considered to choose different copula function in the data model step according to characteristic of shape parameter value on each cluster.

View full abstract

Download PDF (404K)
Adjusting Correlation Matrices Via a Bayesian Approach(Session 6C(IASC-ARS))

Philip L. H. YU, W. K. LI, F. C. NG

Article type: Article
Pages 185-188
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_185

CONFERENCE PROCEEDINGS FREE ACCESS

Download PDF (345K)
Classification trees with interaction detecting variable selection and MSE reducing split point selection(Session 6C(IASC-ARS))

Ahhyoun Kim i, Hyunjoong Kim

Article type: Article
Pages 189-192
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_189

CONFERENCE PROCEEDINGS FREE ACCESS

Download PDF (328K)
THE POISSON-HURWITZ-LERCH ZETA DISTRIBUTION(Session 6C(IASC-ARS))

Kian Wah Liew, Seng Huat Ong

Article type: Article
Pages 193-194
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_193

CONFERENCE PROCEEDINGS FREE ACCESS

Download PDF (170K)
Investigation of Multiple Pathways for Taiwan Liver Cancer(Session 6C(IASC-ARS))

Meng-Yuan Tsai, Henry H. S. Lu, Wai Y. Tan

Article type: Article
Pages 195-198
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_195

CONFERENCE PROCEEDINGS FREE ACCESS

Show abstractHide abstract

Liver cancer or hepatic cancer is a predominant carcinoma in developing country, claiming 60,000 deaths each year worldwide. In Taiwan, it has been on the top, in terms of the death cause in carcinoma with 10,000 deaths each year. In general, hepatocellular carcinoma is raised from two mechanisms: chromosomal instability and micro-satellite instability. However, pediatric hepatic carcinoma can be caused directly by the mutation of certain gene. Based on various biological studies, we adjust the stochastic and state space model under consideration of embryo stage. The generalized Bayesian approach is developed to estimate the parameters. We use this model to fit and analyze Taiwan liver cancer data acquired from Bureau of National Health and Insurance. Our results indicate that this model provides the accurate prediction of liver cancer cases and the development of multiple pathways for hepatocellular carcinoma in Taiwan. Consequently, this model provides useful guidance to monitor future cancer cases.

View full abstract

Download PDF (308K)
N-gram spectral analysis on genomes and biological language models(Session 7C(IASC-ARS))

Yong Kheng GOH, Foo Weng LIM, Yean Ling LEO

Article type: Article
Pages 199-202
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_199

CONFERENCE PROCEEDINGS FREE ACCESS

Show abstractHide abstract

It has been suggested previously that genomic sequences show characteristics typical of natural-language texts such as characteristics "signature" word usage. Thus, the algorithms from natural-language processing may therefore be used in genomic sequences as a mean of clustering and building of phylogenetic tree. Following this approach of natural-language processing, statistical N-gram analysis has been applied for comparative analysis of whole genome sequences of several organisms. It could be shown that a few particular DNA N-grams are found in abundance in one organism but occurring very rarely in other organisms, thereby serving as genome signatures. By counting the occurrences of different N-grams, one can define signature vectors of a genetic text, such as contrast value and usage departure. From the contrast value vectors, phylogenetic trees could be built by using linguistic similarity measures like correlation.

View full abstract

Download PDF (419K)
Testing for genetic association incorporating imprinting effects(Session 7C(IASC-ARS))

Fan Xia, Ji-Yuan Zhou, Wing Kam Fung

Article type: Article
Pages 203-206
Published: November 01, 2012
Released on J-STAGE: July 15, 2017

DOIhttps://doi.org/10.20551/jscssymo.26.0_203

CONFERENCE PROCEEDINGS FREE ACCESS

Show abstractHide abstract

Currently, the increasing interest of the influence of parental imprinting on heritability indicates the importance of incorporating imprinting effects into the mapping of association variants. In this paper, we extend the transmission disequilibrium test (TDT) to incorporate imprinting effects and develop new test statistics in a two-stage framework for association studies. Our test statistics enjoy the nature of familybased designs that need no assumption of Hardy-Weinberg equilibrium. Also, the proposed methods accommodate families with one or more affected children. In the simulation study, we verify the validity of the proposed test statistics under various scenarios, and compare the powers of the proposed statistics with some existing test statistics. It is shown that our methods greatly improve the power for detecting association in the presence of imprinting effects.

View full abstract

Download PDF (334K)

Register with J-STAGE for free!