瀏覽人次: 11753

講者名單及講題

陸方研討會會議手冊

開幕演講人:黃文璋 (高雄大學退休教授)
題目:海峽兩岸機率統計交流25年
摘要
25年前,1996年,在兩岸機率與統計界齊心合作下,第一屆海峽兩岸統計學研討會,在位於台灣高雄的中山大學舉行。會議圓滿成功,並開啟了兩岸機率統計界的交流活動。之後,陸續有兩岸大學統計相關系所的互訪,及聯合舉行會議等。而海峽兩岸機率與統計學術研討會的第二屆、第三屆、…,輪流在兩岸各地舉行,至今已有十一屆。其間雖有若干屆因故延後舉行,惟在兩岸機率與統計界,皆具高度誠意下,本研討會一直持續不斷。25年來,海峽兩岸會議對加強雙方機率統計之互動,及促進彼此了解,發揮極大功能。在本演講中,我們將回顧四分之一世紀前,舉辦第一屆海峽兩岸統計學研討會之經過,並略述其後幾屆的舉辦情形。我們也會對未來兩岸機率與統計之交流,提出若干建議。
經歷
  • 高雄大學副校長(2004年8月31日~2006年10月31日)
  • 中華機率統計學會理事長(2001年4月~2004年4月)
  • 主辦第一屆海峽兩岸統計學研討會(1996.7.15~16) 
  • 第二屆海峽兩岸統計與概率學術研討會(蘇州,1999.7.24~25)組織委員會委員
  • 第三屆海峽兩岸統計與概率學術研討會暨2001中華機率統計學會學術研討會學術委員會委員
  • 第五屆泛華統計協會國際會議(香港,2001.8.17~19)榮譽諮詢委員會委員
  • 第四屆海峽兩岸統計與概率學術研討會(蘭州,2004 8.16~18)組織委員會委員
  • 第五屆海峽兩岸統計與概率學術研討會(新竹國家衛生研究院,2006.7.29~30)籌備委員會委員
  • 第七屆海峽兩岸機率與統計研討會暨2010年中華機率統計學會及學術研討會(花蓮東華大學,2010.5.1~2)議程委員會委員
  • 第八屆海峽兩岸機率與統計研討會(大陸哈爾濱,2012.8.14~16)籌備委員會委員
  • 第九屆海峽兩岸機率與統計學術研討會(台中逢甲大學、國立中興大學,2014.5.16~18)學術指導委員會委員
有關黃教授其他的重要學術成就請參考以下網站:
http://www.stat.nuk.edu.tw/huangwj/cindex.htm

Keynote Speaker: 盧鴻興 (陽明交通大學統計研究所教授)
題目:人工智慧輔助門診的統計學習
摘要
這項研究報告了與台北榮民總醫院聯合開發的AI輔助門診。針對特定的臨床應用,討論了通過醫學圖像的多種模式使用深度學習技術的計算機輔助診斷系統的設計。研究了相關問題、以集成統計模型、計算演算法和領域知識。 總結了當前的發展,並提出了未來的潛在研究。
經歷
盧鴻興教授學術成就卓越, 主要研究領域包括: Statistics, Imagine Science, Bioinformatics, Data Science, Machine Learning and Scientific Computation, 並有眾多論文發表在相關領域的頂尖期刊, 例如: JASA, Statistica Sinica, JCGS, Bioinformatics, IEEE Trans. Medical Image, Biostatistics, PLoS ONE, etc. 盧教授擔任過許多重要期刊的(副)編輯 (如: WIREs Computational Statistics, JASA, J. Applied Mathematics, Statistica Sinica, Journal of Data Science, Handbook of Statistical Bioinformatics, International Journal of Systems and Synthetic Biology, etc), 也是許多重要學術組織的榮譽會員, 並曾經擔任過許多重要的行政職位 (如: Elected member of ISI, Principal fellow of Higher Education Academy (PFHEA), Board of Directors for ICSA, Vice President for NCTU Academic Affairs, Dean for NCTU College of Science, Director of NCTU Big Data Research Center, etc).

有關盧教授其他的重要學術成就請參考以下網站:
https://www.stat.nctu.edu.tw/zh_tw/members/teacher/-%E7%9B%A7%E9%B4%BB%E8%88%88-58192428

 
Session 1-5演講者 (依場次安排)
Session 1 (時間序列及應用:主持人:黄子銘副教授)
演講人1陳婉淑 (逢甲大學統計系)
題目Bayesian estimation of realized GARCH-type models with application to financial tail risk management
摘要
Advances in the various realized GARCH models have proven effective in taking account of the bias in realized volatility (RV) introduced by microstructure noise and non-trading hours. They have been extended into nonlinear or long-memory patterns, including the realized exponential GARCH (EGARCH), realized heterogeneous autoregressive GARCH (HAR-GARCH), and realized threshold GARCH (TGARCH) models. These models with skew Student's t-distribution are applied to quantile forecasts such as Value-at-Risk and expected shortfall of financial returns as well as volatility forecasting. Parameter estimation and quantile forecasting are built on Bayesian Markov chain Monte Carlo sampling methods. Backtesting measures are presented for both Value-at-Risk and expected shortfall forecasts and employ two loss functions to assess volatility forecasts. Results taken from the S&P500 in the U.S. market with approximately 5-year out-of-sample periods covering the COVID-19 pandemic period are reported as follows: (1) The realized HAR-GARCH model performs best in respect of violation rates and expected shortfall at the 1% and 5% significance levels. (2) The realized EGARCH model performs best with regard to volatility forecasts.

演講人2黃士峰 (高雄大學統計研究所)
題目 A Network Autoregressive Model with GARCH Effects and its Applications
摘要
This study proposes a network autoregressive model with GARCH effects, denoted by NAR-GARCH, to depict the return dynamics of stock market indices. A GARCH filter is employed to remove the GARCH effects of each index marginally and the NAR model with the Granger causality test and Pearson's correlation test with sharp price movements is used to capture the joint effects caused by other indices with the most updated market information. The NAR-GARCH model is designed to depict the joint effects of multidimensional time series in an easy-to-implement and effective way. The returns of 20 global stock indices from 2006 to 2020 are employed for our empirical investigation. The numerical results reveal that the NAR-GARCH model has satisfactory performances in both fitting and prediction for the 20 stock indices, especially when a market index has strong upward or downward movements.

演講人3林良靖 (成功大學統計系)
題目 Symbolic interval-valued time series models
摘要
This study considers interval-valued time series data. To characterize such data, we propose an auto-interval-regressive moving average (AIRMA) model using the order statistics from normal distributions. Furthermore, to better capture the heteroscedasticity in volatility, we design a generalized heteroscedastic volatility AIR (GHVAIR) model. We derive the likelihood functions of the AIRMA and GHVAIR models to obtain the maximum likelihood estimator. Monte Carlo simulations are then conducted to evaluate our methods of estimation and confirm their validity. A real data example from the PM2.5 levels on south Taiwan is used to demonstrate our method.

演講人4鄭宇翔 (世新大學財務金融學系)
題目非線性因果關係檢定
摘要
因果關係檢定是一種在經濟及財務領域中用來檢查數列關係的方法,其中最常被使用的為線性架構的Granger因果關係檢定。為了發現數列間可能存在的更一般性之因果關係,許多非線性因果關係檢定已被提出。在此演講中將考慮基於B-spline架構下的Granger因果關係,並探討模型中參數設定之議題,如B-spline節點的選取等問題。最後此研究將使用數種不同的因果關係檢定來分析多國股價之間的因果關係。
 
Session 2(空間統計與存活分析;主持人:黃佳慧副教授)
演講人1陳春樹 (中央統研所)
題目Model selection with an anisotropic nested spatial correlation structure
摘要
In spatial regression analysis, a suitable specification of the mean regression model is crucial for unbiased analysis. Suitably account for the underlying spatial correlation structure of the response variables is also an important issue. Here, we focus on selection of an appropriate mean model in spatial regression analysis under a general anisotropic nested spatial correlation structure. We propose a distribution-free model selection criterion which is an estimate of the weighted mean squared error based on assumptions only for the first two moments of the response data. The simulations under the settings of covariate selection reveal that the proposed criterion performs well for covariate selection in the mean model regardless of the underlying spatial correlation structure is nested, non-nested, isotropic, or anisotropic. Also, the proposed criterion accommodates both continuous and count response data. Finally, a real data example regarding the fine particulate matter concentration is also analyzed for illustration.

演講人2楊洪鼎 (高雄大學統計研究所)
題目A predicting perspective variable selection for the spatial regression model under the presence of spatial confounding
摘要
The spatial regression model has been wildly used in analyzing spatially referenced data, which includes observed covariates and unobserved spatial random effects. When spatial confounding exists, it has been pointed out that parameter estimation and spatial prediction are unreliable. To overcome the issue, we propose a procedure of selecting covariates for spatial regression. We first introduce an adjusted estimation method of regression coefficients and the consequent spatial predictor. Then, we discover a generalized conditional Akaike information criterion to select a subset of covariates, resulting in variable selection and spatial prediction that are satisfactory. Statistical inferences of the proposed methodology are justified theoretically and numerically. This is a joint work with Yung-Huei Chiou and Chun-Shu Chen.

演講人3黃佳慧 (政治大學統計系)
題目Joint Analysis of Dependent Count Data with Excessive Zero
摘要
This work is motivated by an obstructive sleep apnea study in Taiwan, where patients may experience breathing stops for brief periods of time during sleep. Individuals need to take one night at the laboratory and polysomnography (PSG)  is used to measure their sleep variables, such as apnea-hypopnea index (AHI) or respiratory disturbance index (RDI). With this sleep study, there were two types of sleep positions recorded and some observed lateral RDI had a substantial portion of zeros. To this aim, we propose a Bayesian framework to model marginal distribution of event rates and use copula functions to analyze the dependent structure between the sleeping position and position-specific RDI. We develop estimating equations for the model parameters via marginal approach and EM algorithm to obtain estimate of the association parameter. Simulations are conducted to demonstrate the finite sample properties of the proposed analysis, and a real application is reported to illustrate its utility.

演講人4陳政輝 (政治大學應用數學系)
題目Modeling chronic HBV infections with survival probability metrics
摘要
慢性病的病程常以Markov過程建模。由於慢性病的病程很長,通常很難只由一個臨床研究觀察到整個過程,因此,模型參數多是基於不同的短期臨床研究整合得出。儘管這種整合方法為慢性病的病程提供了基本全貌,但在深入分析下可能會導致不切實際的結果。例如,如果沒有仔細校準,模型計算出的患者壽命可能比一般人還長。這樣的結果通常是由於自然死亡率的影響沒有很好地被這些短期研究考慮。學者Beck Pauker 提出一種將自然死亡率整合到病程模型的方法。然而,他們的方法隱含了自然死亡率僅影響轉變到死亡狀態和保持在初始狀態兩路徑的假設,未解釋為什麼只有這兩種類型的路徑會受到自然死亡率的影響。從實際情況來看,無論患者經歷何種狀態轉變,都會面臨自然死亡風險。基於這一觀察,我們提出了一種新的整合方法。假設自然死亡率獨立地影響Markov模型中所有狀態轉換的路徑。這擴展了 Beck Pauker 的想法,也使他們的方法更加合理。我們提出的方法被用於考慮慢性B型肝炎病毒 (HBV) 感染的病程分析。數值結果表明,此方法可以計計算患者的預期壽命與生存概率。此外,首次事件發生時間(first hitting time)的概念被用以推導患者在指定時間內會經歷危急醫療狀態的概率,提供了慢性B型肝炎患者有價值的醫療訊息。我們也討論該方法的可能擴展,例如,所得模型在不同國家的應用以及如何將患者的風險因子引入模型,提供進一步的研究方向。
 
Session 3(生物資訊與應用統計;主持人:吳漢銘副教授)
演講人1葉倚任 (高雄師範大學數學系)
題目Arteriovenous Graft Failure Detection in Hemodialysis
摘要
The fast-growing prevalence of end stage renal disease (ESRD) leads to an increasing burden of population requiring dialysis worldwide. Specifically, patients who choose to receive hemodialysis (HD) face the problem of maintaining their arteriovenous accesses. Unfortunately, occurrence of stenosis and clots is not uncommon, especially in arteriovenous grafts (AVG). Blood flow through grafts contains a number of information that can help prevent these circumstances and improve graft longevity. The aim of this study is to develop a portable recording device that detects stenosis by extracting information from blood flow sounds.

演講人2蘇家玉 (臺北醫學大學醫學資訊研究所)
題目An Infodemiology Study of Internet Search Trends in Spatially Clustered COVID-19 Areas in the United States
摘要
Objective: Incorporating spatial analyses and online health information queries may be beneficial in understanding the role of Google relative search volume (RSV) data as a secondary public health surveillance tool during pandemics. Hence, this study identified COVID-19 clustering and defined the predictability performance of Google RSV models in clustered and non-clustered areas in the United States.
Methods: Getis-Ord General and local G statistics were used to identify monthly clustering patterns. Monthly country- and state-level correlations between new daily COVID-19 cases and Google RSVs were assessed using Spearman’s rank correlation coefficients and Poisson regression models for period of January to December 2020.
Results: Huge clusters involving multiple states were found, which resulted from varied control measures in each state. This demonstrates the importance of state-to-state coordination in implementing control measures to tackle the spread of outbreaks. Variability in Google RSV model performances was found among states and time periods, possibly suggesting the necessity of utilizing different frameworks for Google RSV data in each state. Moreover, the sign of a correlation can be utilized to understand public responses to control and preventive measures, as well as in communicating risk.
 
演講人3馬瀰嘉 (成功大學統計系)
題目發現重要代謝物數據特徵的視覺化統計方法
摘要
本研究目的是使用統計學的視覺化方法發現代謝物數據有顯著差異的特徵,以確定特定疾病在臨床上有用的生物標誌物。在大多數醫學研究上通常使用兩個常見的圖: 火山圖 (volcano plot)和 S圖 (S plot)。火山圖通常使用 t 檢定(或無母數Mann Whitney 檢定方法)的p值當作縱軸,兩組平均數的比值當作橫軸來比較兩組代謝物實驗數據的結果。S圖分別使用每個代謝物特徵和其第一主成分的共變異數和相關係數當作橫軸和縱軸來作圖。但是,過去這兩種圖只使用在兩組代謝物的實驗數據上,以找出顯著的代謝物特徵。本研究放寬火山圖兩組的條件並合併S圖,提出新的視覺化統計圖。我們將火山圖的橫軸改成變異數分析(或非參數 Kruskal Wallis 檢定方法)的p值,縱軸改成每個代謝物特徵和其第一主成分的相關係數。另外,本研究使用兩組實驗數據來說明代謝物的鑑定。研究結果顯示,本研究提出的方法更容易發現重要的代謝物特徵。在火山圖和S圖兩組實驗數據中發現顯著的代謝物特徵,利用本研究提出的方法也都能發現。對於六組實驗數據,透過機器學習方法改良的圖形覆蓋率最高。因此,建議當實驗是三組以上時,可以利用本研究的視覺化統計方法來發現重要代謝物數據特徵。

演講人4陳蔓樺 (淡江大學統計系)
題目門檻回歸模型的競爭風險輔助分配

摘要
門檻回歸模型能處理比例風險和非比例風險的問題並做為Cox比例風險模型的一個替代模型。我們將研究重點放在競爭風險資料在門檻回歸模型下討論,並引入維納過程方法來進一步確認cause-specific hazard、subdistribution hazard、cumulative incidence function (CIF) 之間的關係。我們模擬了 COVID-19 患者臨床試驗的數據,模擬最近的涉及使用羥氯喹、瑞德西韋和恢復期血漿治療的試驗例子。 我們利用門檻回歸模型進一步分析此模擬資料在住院死亡率和康復之間的競爭風險關係。
 
Session 4(多變量分析與機器學習;主持人:薛慧敏教授)
演講人1林宗儀 (中興大學應用數學系)
題目Mixtures of Factor Analyzers with Covariates for Clustering Multiply Censored Dependent Variables
摘要
Censored data arise frequently in diverse applications in which observations to be measured may be subject to some upper and lower detection limits due to the restriction of experimental apparatus such that they are not exactly quantifiable. Mixtures of factor analyzers with censored data (MFAC) have been recently proposed for model-based density estimation and clustering of high-dimensional data under the presence of censored observations. We consider an extended version of MFAC with covariates to accommodate multiply censored dependent variables and develop two analytically feasible EM-type algorithms for computing maximum likelihood (ML) estimates of model parameters with closed-form expressions. Moreover, we provide an information-based method to compute asymptotic standard errors of mixing proportions and regression coefficients. The utility and performance of our proposed methodologies are illustrated through several simulated experiments and real data examples.

演講人2呂恆輝 (東海大學統計系)
題目Interpretable, predictive spatio-temporal models via enhanced Pairwise Directions Estimation
摘要
This article concerns the predictive modeling for spatio-temporal data as well as model interpretation using data information in space and time. Intrinsically, we develop a novel approach based on dimension reduction for such data in order to capture nonlinear mean structures without requiring a prespecified parametric model. In addition to prediction as a common interest, this approach focuses more on the exploration of geometric information in the data. The method of Pairwise Directions Estimation (PDE) is incorporated in our approach to implement the data-driven function searching of spatial structures and temporal patterns, useful in exploring data trends. The benefit of using geometrical information from the method of PDE is highlighted. We further enhance PDE, referring to it as PDE+, by using resolution adaptive fixed rank kriging to estimate the random effects not explained in the mean structures. Our proposal can not only produce more accurate and explainable prediction, but also increase the computation efficiency for model building. Several simulation examples are conducted and comparisons are made with four existing methods. The results demonstrate that the proposed PDE+ method is very useful for exploring and interpreting the patterns of trend for spatio-temporal data. Illustrative applications to two real datasets are also presented.

演講人3王婉倫 (逢甲大學統計系)
題目A Bayesian Approach to Multivariate Linear Mixed Models with Censored and Missing Responses
摘要
Multivariate longitudinal data usually exhibit complex features such as the presence of censored responses due to detection limits of the assay and unavoidable missing values arising when participants make irregular visits that lead to intermittently recorded characteristics. A generalization of the multivariate linear mixed model constructed by taking into account impacts of censored and intermittent missing responses simultaneously, which is named as the MLMM-CM, has been recently proposed for more precisely analyzing such kinds of data. This paper aims at presenting a fully Bayesian approach to the MLMM-CM for addressing the uncertainties of censored and missing responses as well as unknown parameters. Bayesian computational techniques based on the inverse Bayes formulas (IBF) coupled with the Gibbs scheme are developed for carrying out posterior inference of the model. The proposed methodology is illustrated through a simulation study and a real-data example from the Adult AIDS Clinical Trials Group 388 study. Numerical results show empirically that the proposed Bayesian methodology performs satisfactorily and offers reliable posterior inference.

演講人4郁方 (政治大學資訊管理系)
題目HiSeqGAN: Hierarchical Sequence Synthesis and Prediction
摘要
High-dimensional data sequences constantly appear in practice. State-of-the-art models such as recurrent neural networks suffer prediction accuracy from complex relations among values of attributes. Clustering data based on their attribute value similarity provides a way to abstract data to clusters in a lower dimension, which could be integrated to improve prediction performance. In this talk, we propose HiSeqGAN, a new generator to synthesize and predict sequences of high dimensional data that are structured in a hierarchy. By relabeling a sample with the cluster that it falls to, we are able to use the GHSOM map to abstract high-dimensional data with clusters in an hierarchical structure that represents its attribute value similarity. We converse the clusters to the coordinate vectors with our hierarchical data encoding algorithm and replace the loss function with maximizing cosine similarity in the SeqGAN model to synthesize cluster sequences. Using the synthesized sequences, we are able to achieve better performance on high-dimension data training and prediction compared to the state-of-the-art models. Specifically, HiSeqGAN adopts a GAN-like learning mechanism. Our generator model is updated by employing a policy gradient and Monte Carlo search, where the final reward signal is provided by the discriminator based on coordinate sequence cosine similarity and is passed back to the intermediate action value. To predict next tokens of training sequences, we use the trained generator to generate sequences that have more periods, and the extended tokens are used as our prediction on sequences that have similar prefix. We have collected two-year real transactions of a semiconductor component distributor to conduct our experiments. The preliminary results show that we can achieve better accuracy compared to standard approaches for sequence prediction on clusters and attribute values, e.g., weekly demands.
 
Session 5(工業統計與應用;主持人:翁久幸教授)
演講人1楊素芬 (政治大學統計系)
題目A Loss Control Chart for Monitoring Multiple Bottle Cover Weights
摘要
The cosmetic bottle cover weights are collected from a filling machine with multiple identical filling heads every four hours. The cover weights are a critical quality variable of the bottles. The weights variation from each filling head are large, and the weights distributions from each filling head are correlated but not identical. To monitor whether the processes are in-control or out-of-control, the EWMA loss control charts are proposed to demonstrate the in-control samples and detect the out-of-control samples. We find that the proposed EWMA loss control charts perform well.

演講人2黃怡婷 (台北大學統計系)
題目An Assessment System for Predicting Suitable Range of Heel Height Using Regression
摘要
Wearing high-heeled shoes that exceed the tolerant height might cause severe injuries for females. Such injuries include musculoskeletal pain, osteoarthritis and hallux valgus. The most tolerant height of high-heeled shoes for females can be determined by touching calcaneus to find the deformation of calcaneal varus. Nevertheless, it is not very convenient to visit physician just to find the tolerant height. A novel system was developed to obtain the plantar pressure. 100 females satisfying the inclusion criteria were enrolled in this study. Each participant was asked to stand on the platform imbedding with the 21 sensors to measure the plantar pressure. By adjusting the platform, the plantar pressure for 21 heights was collected. Furthermore, the exact tolerant height of high-heeled shoes of the participant was determined by the physician. Using the regression model, this study included three modules to analyze the data for each feet. We found that the predicted power for the model using the plantar pressure collected from two extreme heights achieves 0.82 and 0.77 for the right and left feet, respectively.

演講人3蔡志群 (淡江大學數學系)
題目Lamination Design of Photovoltaic Solar Panels
摘要
Solar power is inexhaustible and has become the most appreciative choice in the world. With development stage of solar modules, solar panels are conducted by the relevant reliability tests to ensure long lifetime and power generation efficiency. During the lamination process of solar modules, the performance of the solar panels has been greatly relevant with the degree of crosslinking for EVA sheet. The degree of crosslinking for EVA sheet is obtained by using the extraction method to measure the gel content of EVA sheet. Motivated by lamination tests on solar panels, this study first constructed the statistical model with extreme value residuals to describe the relationship between the degree of cross-linking for EVA sheet and lamination time. Then, under the specification upper and lower limits of the degree of cross-linking for EVA sheet, the optimal lamination time of solar panels will be derived, and the optimal sample allocation for measuring EVA sheets destructively will be addressed.

演講人4王義富 (中正大學數學系)
題目Degradation Analysis on Trend Gamma Process
摘要
Manufactures always face a big challenge to obtain the sufficient reliability information for high quality products when only a relatively short time period is available for an internal life testing. Fortunately, if quality characteristics (QC) exists, whose degradation over time can be related to reliability, an approach is suggested to collect sufficient degradation data for estimating the product’s lifetime distribution more accurately. In numerical applications, the gamma process (GP) is commonly used when the degradation path is strictly increasing. However, in some circumstances the gamma process is not able to successfully capture the degradation path. In order to tackle this problem, Lawless and Crowder (2004) considered the random effect into the GP model and analyzed the fatigue crack growth data taken from Meeker and Escobar (1998). In this paper, motivated by the effect of trend function proposed by Lindqvist et al. (2003), we propose a trend gamma process (TGP) which is integrated the merits of trend function into a GP model, as an alternative approach to overcome this obstacle. The proposed TGP model is a generalized formulation of GP model, and it is attempted to transform a monotonic stochastic process into a gamma process. The results based on simulation studies and real data examples illustrate our proposed TGP model is suitable for widely use when the degradation path is monotonically increasing.