P
ROGRAMME AND
A
BSTRACTS
1st International Conference on
Econometrics and Statistics (EcoSta 2017)
http://cmstatistics.org/EcoSta2017
Hong Kong University of Science and Technology
15 – 17 June 2017
ISBN: 978-9963-2227-2-8
c
2017 - ECOSTA Econometrics and Statistics
Technical Editors: Gil Gonzalez-Rodriguez and Marc Hofmann.
All rights reserved. No part of this book may be reproduced, stored in a retrieval system, or transmitted, in any other form or by any means without the prior permission from the publisher.
Co-chairs:
Ana Colubi, Erricos J. Kontoghiorghes, Tsung-I Lin, Yasuhiro Omori, Byeong Park and Mike K.P. So.
Scientific Programme Committee:
Alessandra Amendola, Eric Beutner, Monica Billio, Cathy W.S. Chen, Ray-bin Chen, Ming-Yen Cheng,
Jeng-Min Chiou, Sung Nok Chiu, Taeryon Choi, Bertrand Clarke, Aurore Delaigle, Jean-Marie Dufour,
Yang Feng, Alain Hecq, Inchi Hu, Salvatore Ingrassia, Yongdai Kim, Robert Kohn, Carlos Lamarche, Degui
Li, WK Li, Yi Li, Zudi Lu, Geoff McLachlan, Samuel Mueller, Marc Paolella, Tommaso Proietti, Artem
Prokhorov, Igor Pruenster, Jeroen Rombouts, Huiyan Sang, Tak Kuen Siu, Xinyuan Song, Mark Steel,
Jianguo Sun, Nobuhiko Terui, Alan Wan, Lan Wang, Toshiaki Watanabe, Yoon-Jae Whang, Heung Wong,
Valentin Zelenyuk, Helen Zhang, Ping-Shou Zhong and Hongtu Zhu.
Local Organizing Committee:
Dear Colleagues,
It is a great pleasure to welcome you to the 1st International Conference on Econometrics and Statistics (EcoSta
2017). The Conference is co-organized by the Working Group on Computational and Methodological Statistics
(CMStatistics), the Network of Computational and Financial Econometrics (CFENetwork), and the Hong Kong
University of Science and Technology (HKUST) Business School.
The aim is for the conference to become a leading meeting in econometrics, statistics and their applications.
The EcoSta 2017 consists of almost 150 sessions, four keynote talks, four invited sessions, and about 600
presen-tations. There are over 650 participants. This is quite impressive for a first edition of a conference. It is indeed
promising that the EcoSta conference will become a successful medium for the dissemination of high quality
re-search in Econometrics and Statistics, and facilitate networking.
The Co-chairs observed that the collective effort of the scientific program committee, session organizers, and local
organizing committee has produced a programme that spans all the areas of econometrics and statistics. The
HKUST provides excellent facilities and a fantastic environment with an astonishing scenery on the outskirts of
Hong Kong. The local host, volunteers, and sponsoring universities have substantially contributed through their
effort to the successful organization of the conference. We thank them all for their support. Particularly we express
our sincere appreciation to the host and main sponsor, the HKUST Business School.
It is hoped that the quality of both the scientific programme and the HKUST will provide the participants with a
productive, stimulating conference, and an enjoyable stay in Honk Kong.
The Elsevier journals of Econometrics and Statistics (EcoSta) and Computational Statistics & Data Analysis
(CSDA) are associated with CFEnetwork, CMStatistics, and the EcoSta 2017 conference. The participants are
encouraged to submit their papers to special or regular peer-reviewed issues of EcoSta and CSDA, and to join the
networks.
Finally, we are happy to announce that the 2nd International Conference on Econometrics and Statistics (EcoSta
2018) will take place at the City University of Hong Kong from Tuesday 19 to Thursday 21 of June 2018. You are
invited to participate actively in these events.
Ana Colubi, Erricos J. Kontoghiorghes and Mike K.P. So
on behalf of the Co-Chairs
CMStatistics: ERCIM Working Group on
COMPUTATIONAL AND METHODOLOGICAL STATISTICS
http://www.cmstatistics.orgThe working group (WG) CMStatistics comprises a number of specialized teams in various research areas of computational and methodological statistics. The teams act autonomously within the framework of the WG in order to promote their own research agenda. Their activities are endorsed by the WG. They submit research proposals, organize sessions, tracks and tutorials during the annual WG meetings and edit journal special issues. The Econometrics and Statistics (EcoSta) and Computational Statistics & Data Analysis (CSDA) are the official journals of the CMStatistics.
Specialized teams
Currently the ERCIM WG has over 1650 members and the following specialized teams BM: Bayesian Methodology
CODA: Complex data structures and Object Data Analysis CPEP: Component-based methods for Predictive and
Ex-ploratory Path modeling
DMC: Dependence Models and Copulas DOE: Design Of Experiments
EF: Econometrics and Finance
GCS: General Computational Statistics WG CMStatistics GMS: General Methodological Statistics WG CMStatistics GOF: Goodness-of-Fit and Change-Point Problems
HDS: High-Dimensional Statistics
ISDA: Imprecision in Statistical Data Analysis LVSEM: Latent Variable and Structural Equation Models
MCS: Matrix Computations and Statistics
MM: Mixture Models
MSW: Multi-Set and multi-Way models NPS: Non-Parametric Statistics
OHEM: Optimization Heuristics in Estimation and Modelling RACDS: Robust Analysis of Complex Data Sets
SAE: Small Area Estimation
SAET: Statistical Analysis of Event Times SAS: Statistical Algorithms and Software SEA: Statistics of Extremes and Applications SFD: Statistics for Functional Data
SL: Statistical Learning
SSEF: Statistical Signal Extraction and Filtering TSMC: Times Series Modelling and Computation
You are encouraged to become a member of the WG. For further information please contact the Chairs of the specialized groups (see the WG’s website), or by email at info@cmstatistics.org.
CFEnetwork
COMPUTATIONAL AND FINANCIAL ECONOMETRICS
http://www.CFEnetwork.orgThe Computational and Financial Econometrics (CFEnetwork) comprises a number of specialized teams in various research areas of theoretical and applied econometrics, financial econometrics and computation, and empirical finance. The teams contribute to the activities of the network by organizing sessions, tracks and tutorials during the annual CFEnetwork meetings, and by submitting research proposals. Furthermore the teams edit special issues currently published under the Annals of CFE. The Econometrics and Statistics (EcoSta) is the official journal of the CFEnetwork.
Specialized teams
Currently the CFEnetwork has over 1000 members and the following specialized teams AE: Applied Econometrics
BE: Bayesian Econometrics BM: Bootstrap Methods
CE: Computational Econometrics
ET: Econometric Theory FA: Financial Applications FE: Financial Econometrics TSE: Time Series Econometrics
You are encouraged to become a member of the CFEnetwork. For further information please see the website or contact by email at info@cfenetwork.org.
SCHEDULE
2017-06-14 2017-06-15 A - Keynote EcoSta2017 09:00 - 09:50 B EcoSta2017 10:25 - 12:30 C EcoSta2017 14:00 - 15:40 D EcoSta2017 16:10 - 17:25 E - Keynote EcoSta2017 17:40 - 18:30 2017-06-16 F EcoSta2017 08:30 - 09:50 G EcoSta2017 10:20 - 12:25 H - Keynote EcoSta2017 14:00 - 14:50 I EcoSta2017 15:00 - 16:40 J EcoSta2017 17:10 - 18:50 2017-06-17 K EcoSta2017 08:30 - 09:50 L EcoSta2017 10:15 - 11:30 M - Keynote EcoSta2017 11:40 - 12:30 N EcoSta2017 14:00 - 15:40 O EcoSta2017 16:10 - 17:50Registration & Ice Breaker
17:30 - 19:00 Opening, 08:45 - 09:00 Coffee Break 09:50 - 10:25 Lunch Break 12:30 - 14:00 Coffee Break 15:40 - 16:10 Welcome Reception 18:45 - 20:15 Coffee Break 09:50 - 10:25 Lunch Break 12:30 - 14:00 Coffee Break 16:40 - 17:10 Conference Dinner 19:45 - 22:15 Coffee Break, 09:50 - 10:15 Lunch Break 12:30 - 14:00 Coffee Break 15:40 - 16:10 Closing, 17:55 - 18:10 Closing Drink, 18:15 - 18:40
MEETINGS AND SOCIAL EVENTS SPECIAL MEETINGSby invitation to group members
• The Econometrics and Statistics (EcoSta) Editorial Board meeting will take place on Saturday 17th of June 2017, 09:50-10:50, Room LSK1032. The meeting is by invitation only.
• The Econometrics and Statistics (EcoSta) Editorial Board dinner will take place on Saturday 17th of June 2017, 20:00-22:30. A minibus will depart from the conference venue at 19:20. The dinner is by invitation only.
SOCIAL EVENTS
• Registration & Ice breaker, Wednesday 14th of June 2017, from 17:30 - 19:00. The Ice breaker is open to all registrants and accompanying persons. It will take place at the Hall of the ground floor (see map at page IX).
• The coffee breaks will take place at the Conference Lodge (see map at page VIII). You must have your conference badge in order to attend the coffee breaks.
• Welcome Reception, Thursday 15th of June 2017, from 18:45 - 20:15. The Welcome Reception is open to all registrants and accompanying persons who have purchased a reception ticket. It will take place at the Lo Ka Chung University Center (see maps at page VIII). Conference registrants must bring their conference badge and ticket and any accompanying persons shall bring their reception tickets in order to attend the reception. Preregistration is required due to health and safety reasons, and limited capacity of the venue. Entrance to the reception venue will be strictly allowed only to those who have a ticket. The Welcome Reception is fully sponsored by the HKUST Business School. • Conference Dinner, Friday 16th of June, from 19:45 to 22:15. The conference dinner is optional and registration is required. It will take
place at the Conference Lodge (see map at page VIII). Conference registrants and accompanying persons shall bring their conference dinner tickets in order to attend the conference dinner.
• Conference Buffet Lunches. The conference lunches are optional and registration is required. The Lunch will take place at the Conference Lodge the 15th, 16th and 17th of June 2017. Conference registrants and accompanying persons shall bring their conference lunch tickets in order to attend the conference lunches.
• Conference Lunch box. The conference lunch box is optional and registration is required. The Lunch box will be arranged at the 7th floor, Lounge and Function room (see map at pages VIII and X) of the venue (Lee Shau Kee Business Building) the 15th, 16th and 17th of June 2017. Conference registrants and accompanying persons shall have the corresponding Lunch box ticket in order to attend the lunch each day. • Closing Drink, Saturday 17th of June 2017, from 18:15 - 18:40. The Closing Drink is open to all registrants. It will take place at the Hall of
the ground floor (see map at pages VIII and IX).
GENERAL INFORMATION Addresses of venues
• Lee Shau Kee Business Building, The Hong Kong University of Science and Technology (HKUST) Business School, Clear Water Bay, Kowloon Hong Kong.
Registration
The registration will be open from Wednesday afternoon 14th until Saturday 17th June 2017 at the ground floor of the conference venue.
Lecture rooms
The paper presentations will take place at the ground and first floor of the Lee Shau Kee Business Building (see map in page IX). The different rooms are shown in the following floor plans of the venue. We advise that you visit the venue in advance. The opening and keynote talks will take place at the Auditorium LTA – Citi Lecture Theater (see map in page VIII) and the closing talk will take place at room LSKG001 of the venue (see map in page IX).
Presentation instructions
The lecture rooms will be equipped with a PC and a computer projector. The session chairs should obtain copies of the talks on a USB stick before the session starts (use the lecture room as the meeting place), or obtain the talks by email prior to the start of the conference. Presenters must provide the session chair with the files for the presentation in PDF (Acrobat) or PPT (Powerpoint) format on a USB memory stick. This must be done at least ten minutes before each session. Chairs are requested to keep the sessions on schedule. Papers should be presented in the order they are listed in the programme for the convenience of attendees who may wish to go to other rooms mid-session to hear particular papers. In the case of a presenter not attending, please use the extra time for a break or a discussion so that the remaining papers stay on schedule. The PC in the lecture rooms should be used for presentations. An IT technician will be available during the conference and should be contacted in case of problems.
Posters
The poster sessions will take place at the Hall of the ground floor. The posters should be displayed only during their assigned session. The authors will be responsible for placing the posters in the poster panel displays and removing them after the session. The maximum size of the poster is A0.
Internet connection
The information for the wireless Internet connection will be displayed by the registration desk.
Information and messages
You may leave messages for each other on the bulletin board by the registration desks. General information about restaurants, useful numbers, etc. can be obtained from the registration desk.
Exhibitors
Map of the venue and nearby area
Venue: LSK - Ground Floor
PUBLICATION OUTLETS
Econometrics and Statistics (EcoSta)http://www.elsevier.com/locate/ecosta
Econometrics and Statistics (EcoSta), published by Elsevier, is the official journal of the networks Computational and Financial Econometrics (CFENetwork) and Computational and Methodological Statistics (CMStatistics). It publishes research papers in all aspects of econometrics and statistics and comprises two sections:
Part A: Econometrics.Emphasis is given to methodological and theoretical papers containing substantial econometrics derivations or showing a potential of a significant impact in the broad area of econometrics. Topics of interest include the estimation of econometric models and associated inference, model selection, panel data, measurement error, Bayesian methods, and time series analyses. Simulations are considered when they involve an original methodology. Innovative papers in financial econometrics and its applications are considered. The covered topics include portfolio allocation, option pricing, quantitative risk management, systemic risk and market microstructure. Well-founded applied econometric studies that demonstrate the practicality of new procedures and models are of interest as well. Such studies should involve the rigorous application of statistical techniques, including estimation, inference and forecasting. Topics include volatility and risk, credit risk, pricing models, portfolio management, and emerging markets. Innovative contributions in empirical finance and financial data analysis that use advanced statistical methods are encouraged. The results of the submissions should be replicable. Applications consisting only of routine calculations are not of interest to the journal.
Part B: Statistics. Papers providing important original contributions to methodological statistics inspired in applications are considered for this section. Papers dealing, directly or indirectly, with computational and technical elements are particularly encouraged. These cover developments concerning issues of high-dimensionality, re-sampling, dependence, robustness, filtering. In general, the interaction of mathematical methods, numerical implementations and the extra burden of analysing large and/or complex datasets with such methods in different areas such as medicine, epidemiology, biology, psychology, climatology and communication. Innovative algorithmic developments are also of interest, as are the computer programs and the computational environments that implement them as a complement.
The journal consists, preponderantly, of original research. Occasionally, reviews and short papers from experts are published, which may be accompanied by discussions. Special issues and sections within important areas of research are occasionally published. The journal publishes as a supplement the Annals of Computational and Financial Econometrics.
Call For Papers Econometrics and Statistics (EcoSta)
http://www.elsevier.com/locate/ecosta
Papers containing novel components in econometrics and statistics are encouraged to be submitted for publication in special peer-reviewed, or regular issues of the new Elsevier journal Econometrics and Statistics (EcoSta) and its supplement Annals of Computational and Financial Econo-metrics. The Econometrics and Statistics (EcoSta) is inviting submissions for the special issues:
• (Part A: Econometrics) Annals of Computational and Financial Econometrics • (Part A: Econometrics) Special Issue on Forecast combinations.
• (Part A: Econometrics) Special Issue on Risk management.
• (Part B: Statistics) Special Issue on Quantile regression and semiparametric methods. • (Part B: Statistics) Special Issue on Statistics of extremes and applications.
The deadline for paper submissions is the 30th June 2017. Papers should be submitted using the Elsevier Electronic Submission tool EES: http://ees.elsevier.com/ecosta (in the EES please select the appropriate special issue). For further information please consult http://www.cfenetwork.org or http://www.cmstatistics.org.
Call For Papers Computational Statistics & Data Analysis (CSDA)
http://www.elsevier.com/locate/csda
Papers containing strong computational statistics, or substantive data-analytic elements can also be submitted to special peer-reviewed, or regular issues of the journal Computational Statistics & Data Analysis (CSDA). The CSDA is planning for 2017 the following special issues:
• 4th Special Issue on advances in mixture models • Special Issue on Biostatistics
• High-dimensional and functional data analysis
The deadline for paper submissions is the 30th November 2017. Papers should be submitted using the Elsevier Electronic Submission tool EES: http://ees.elsevier.com/csda (in the EES please select the appropriate special issue). Any questions may be directed via email to: csda@dcs.bbk.ac.uk.
Contents
General Information I
Committees . . . III Welcome . . . IV CMStatistics: ERCIM Working Group on Computational and Methodological Statistics . . . V CFEnetwork: Computational and Financial Econometrics . . . V Scientific programme . . . V Meetings and Social events . . . VII Venue, Registration, Social Events, Presentation instructions, Posters and Internet connection . . . VII Map of the venue and nearby area . . . VIII Floor maps . . . VIII Publications outlets of the journals EcoSta and CSDA and Call for papers . . . XI
Keynote Talks 1
Keynote talk 1 (Xuming He, University of Michigan, United States) Thursday 15.06.2017 at 09:00 - 09:50
A statistical tale of subgroup analysis for managerial decision making . . . 1
Keynote talk 2 (Marc Hallin, Universite Libre de Bruxelles, Belgium) Thursday 15.06.2017 at 17:40 - 18:30 Quantile spectral analysis for locally stationary time series . . . 1
Keynote talk 3 (Marc Paolella, University of Zurich, Switzerland) Friday 16.06.2017 at 14:00 - 14:50 Robust normal mixtures for financial portfolio allocation . . . 1
Keynote talk 4 (Michael Pitt, Kings College London, United Kingdom) Saturday 17.06.2017 at 11:40 - 12:30 Recent developments in Bayesian inference for time series . . . 1
Parallel Sessions 2 Parallel Session B – EcoSta2017 (Thursday 15.06.2017 at 10:25 - 12:30) 2 EO116: NEW DEVELOPMENT IN ANALYZING LARGE COMPLEX DATA(Room: LSK1005) . . . 2
EO018: STATISTICAL MODELLING FOR NETWORK DATA(Room: LSK1007) . . . 2
EO136: STATISTICAL METHODS FOR FUNCTIONAL DATA(Room: LSK1011) . . . 3
EO108: MODEL AVERAGING,SELECTION AND SHRINKAGE(Room: LSK1001) . . . 4
EO112: NEW DEVELOPMENTS IN FINANCIAL ECONOMETRICS(Room: LSK1034) . . . 5
EO010: MODELLING FINANCIAL AND INSURANCE RISKS(Room: LSK1027) . . . 5
EO094: ADVANCES IN TIME SERIES ANALYSIS(Room: LSK1033) . . . 6
EO166: APPLIED STATISTICAL MODELING(Room: LSK1032) . . . 7
EO016: ADVANCES IN NONPARAMETRIC METHODS AND APPLICATIONS(Room: LSKG003) . . . 7
EO298: CHANGE POINT ANALYSIS IN A HIGH-DIMENSIONAL SETTING(Room: LSK1010) . . . 8
EO084: MODELLING WITH NON-GAUSSIAN DISTRIBUTIONS(Room: LSKG001) . . . 9
EO080: VARIABLE SELECTION,DIMENSION REDUCTION,AND OUTLIER DETECTION(Room: LSK1014) . . . 9
EO054: STATISTICAL METHODS FOR BIG DATA INTEGRATION(Room: LSK1009) . . . 10
EO206: HIGH DIMENSIONALBAYESIAN TIME SERIES MODELING AND FORECASTING(Room: LSK1003) . . . 11
Parallel Session C – EcoSta2017 (Thursday 15.06.2017 at 14:00 - 15:40) 12 EI301: NON-AND SEMI-PARAMETRIC INFERENCE(Room: LSKG001) . . . 12
EO244: MACRO AND FINANCIAL ECONOMETRICS(Room: LSK1033) . . . 12
EO256: BUSINESS ANALYTICS(Room: LSK1003) . . . 13
EO208: HIGH DIMENSIONAL MATRICES AND NETWORKS(Room: LSK1005) . . . 13
EO305: DATA ANALYTICS AND MACHINE LEARNING METHODS FOR RISK AND INSURANCE(Room: LSKG007) . . . 14
EO152: RECENT ADVANCES IN NONPARAMETRIC INFERENCE (Room: LSK1009) . . . 14
EO154: RECENT ADVANCES IN TIME SERIES ANALYSIS(Room: LSK1034) . . . 15
EO074: ADVANCES IN EXACT AND APPROXIMATEBAYESIAN COMPUTATION (Room: LSK1014) . . . 16
EO178: FACTOR MODELS AND FINANCIAL ECONOMETRICS(Room: LSK1001) . . . 16
EO024: ADVANCES IN CHANGE POINTS,MISSING DATA AND NEURAL NETWORKS(Room: LSK1007) . . . 17
EO144: HIGH DIMENSIONAL INFERENCE FOR COMPLEX DATA(Room: LSK1010) . . . 17
EO176: LARGE SCALE FINANCIAL DATA(Room: LSK1032) . . . 18
EO246: RECENT DEVELOPMENTS IN SUFFICIENT DIMENSION REDUCTION AND GRAPHICAL MODELS(Room: LSK1011) . . . 19
Parallel Session D – EcoSta2017 (Thursday 15.06.2017 at 16:10 - 17:25) 20 EO212: MODELLING AND ESTIMATION IN FINANCIAL TIME SERIES(Room: LSKG007) . . . 20
EO238: EXTREME VALUE MODELING AND RISK ANALYSIS(Room: LSKG001) . . . 20
EO224: MODERN STATISTICAL METHODS FOR COMPLEX DATA(Room: LSK1009) . . . 21
EO020: KSSSESSION: STATISTICAL LEARNING(Room: LSK1010) . . . 21
EO200: HIGH-DIMENSIONAL STATISTICS: TESTING,ESTIMATION AND BEYOND (Room: LSK1014) . . . 21
EO252: RECENT ADVANCES IN COMPLEXLY-STRUCTURED TIME SERIES ANALYSIS(Room: LSK1003) . . . 22
EO132: FINDING GROUP STRUCTURES IN BIOMEDICAL AND HEALTH DATA(Room: LSK1007) . . . 22
EO122: RECENT DEVELOPMENTS ON DYNAMIC TREATMENT REGIMES(Room: LSK1005) . . . 23
EO126: SOME NEW DEVELOPMENT IN COMPLEX SURVIVAL DATA(Room: LSKG003) . . . 23
EO262: NEW DEVELOPMENTS IN BIOMEDICAL RESEARCHI (Room: LSK1001) . . . 24
EO100: NEW DEVELOPMENTS IN EXPERIMENTAL DESIGNS AND INDUSTRIAL STATISTICS(Room: LSK1011) . . . 24
EG165: CONTRIBUTIONS IN SEMI-PARAMETRIC METHODS IN ECONOMETRICS(Room: LSK1034) . . . 25
EG003: CONTRIBUTIONS IN APPLIED ECONOMETRICS(Room: LSK1032) . . . 25
Parallel Session F – EcoSta2017 (Friday 16.06.2017 at 08:30 - 09:50) 26 EO066: RECENT ADVANCE IN TIME SERIES ECONOMETRICS(Room: LSKG003) . . . 26
EO196: BIG DATA AND ITS APPLICATIONS(Room: LSK1014) . . . 26
EG297: CONTRIBUTIONS IN STATISTICAL MODELS WITH APPLICATIONS(Room: LSK1007) . . . 27
EG029: CONTRIBUTIONS IN FORECASTING ECONOMIC AND FINANCIAL TIME SERIES(Room: LSK1034) . . . 27
EG069: CONTRIBUTIONS IN VOLATILITY MODELLING AND FORECASTING(Room: LSK1003) . . . 28
EC288: CONTRIBUTIONS IN ROBUST METHODS(Room: LSK1005) . . . 29
EG013: CONTRIBUTIONS IN HIGH DIMENSIONAL AND COMPLEX DATA ANALYSIS(Room: LSKG001) . . . 29
EC295: CONTRIBUTIONS IN METHODOLOGICAL STATISTICS AND ECONOMETRICS(Room: LSK1010) . . . 30
EC293: CONTRIBUTIONS IN ECONOMETRICS MODELS(Room: LSK1001) . . . 30
EG011: CONTRIBUTIONS IN MODELLING FINANCIAL AND INSURANCE RISKS(Room: LSKG007) . . . 31
EC282: CONTRIBUTIONS IN COMPUTATIONAL AND NUMERICAL METHODS(Room: LSK1009) . . . 31
Parallel Session G – EcoSta2017 (Friday 16.06.2017 at 10:20 - 12:25) 33 EO104: INFERENCE FOR CORRELATED DATA(Room: LSK1014) . . . 33
EO218: ECONOMETRIC METHODS FOR MACROECONOMIC ANALYSIS AND FORECASTING(Room: LSK1001) . . . 33
EO128: NEW DEVELOPMENTS IN SURVIVAL ANALYSIS(Room: LSK1010) . . . 34
EO240: RECENT ADVANCES IN MIXTURE MODELS AND LATENT VARIABLE MODELS(Room: LSK1027) . . . 35
EO158: INSURANCE MODELS WITH DEPENDENCE(Room: LSK1033) . . . 35
EO052: RECENT ADVANCES IN HIGH DIMENSIONAL STATISTICAL INFERENCE(Room: LSK1007) . . . 36
EO234: NEW CHALLENGES IN COMPLEX DATA ANALYSIS(Room: LSKG003) . . . 37
EO140: FINANCIAL VOLATILITY(Room: LSK1032) . . . 37
EO042: LARGE-SCALE,NON-ELLIPTIC PORTFOLIO OPTIMIZATION(Room: LSKG007) . . . 38
EO280: MODELLING FINANCIAL MARKET DYNAMICS(Room: LSK1003) . . . 39
EO250: HIGH DIMENSIONAL PROBLEMS IN ECONOMETRICS(Room: LSK1026) . . . 39
EO264: DESIGN AND ANALYSIS OF COMPLEX EXPERIMENTS: THEORY AND APPLICATIONS(Room: LSK1005) . . . 40
EO148: BAYESIAN NONPARAMETRICS(Room: LSKG001) . . . 41
EO096: STATISTICAL COMPUTING FOR LARGE PANEL DATA(Room: LSK1034) . . . 41
EO190: QUANTILE REGRESSION AND ROBUST METHODS(Room: LSK1009) . . . 42
EO124: RECENT ADVANCES IN LATENT VARIABLE MODELS(Room: LSK1011) . . . 43
EP001: POSTERSESSION(Room: Ground Floor Hall) . . . 43
Parallel Session I – EcoSta2017 (Friday 16.06.2017 at 15:00 - 16:40) 45 EI006: ADVANCES IN SPATIAL STATISTICS(Room: LSKG001) . . . 45
EO022: RECENT ADVANCES ON THE ANALYSIS OF EVENT HISTORY STUDIES(Room: LSK1007) . . . 45
EO303: WAVELETS IN ECONOMICS AND FINANCE(Room: LSK1034) . . . 46
EO012: HIGH DIMENSIONAL AND COMPLEX DATA ANALYSIS(Room: LSK1010) . . . 47
EO230: RECENT CHALLENGES IN GENETIC ASSOCIATION ATUDIES(Room: LSK1011) . . . 47
EO164: NON-AND SEMI-PARAMETRIC METHODS FOR ECONOMICS AND FINANCIAL DATA(Room: LSK1032) . . . 48
EO046: NEW DEVELOPMENTS IN TIME SERIES ANALYSIS (Room: LSKG007) . . . 49
EO216: STATISTICAL INFERENCE FOR HIGH-DIMENSIONAL DATA(Room: LSK1014) . . . 49
EO309: SMART BETA AND QUANTITATIVE INVESTING(Room: LSK1003) . . . 50
EO026: NETWORKS AND CAUSALITY(Room: LSK1001) . . . 50
EO214: INFERENCE AND APPLICATIONS FOR TIME SERIES MODELS(Room: LSK1033) . . . 51
EO248: INTEGRATING BIG AND COMPLEX IMAGING DATA WITH NEW STATISTICAL TOOLS(Room: LSKG003) . . . 51
EO034: RECENT ADVANCES ON HYPOTHESIS TESTING(Room: LSK1009) . . . 52
Parallel Session J – EcoSta2017 (Friday 16.06.2017 at 17:10 - 18:50) 53 EI002: MODERN METHODS FOR COMPLEX FUNCTIONAL AND LONGITUDINAL DATA(Room: LSKG001) . . . 53
EO060: TIME SERIES MODELING AND ITS APPLICATIONS(Room: LSK1001) . . . 53
EO134: FINANCIAL AND RISK MANAGEMENT APPLICATIONS(Room: LSKG007) . . . 54
EO220: NEW METHODS AND APPLICATIONS IN QUANTILE REGRESSION AND BEYOND(Room: LSK1005) . . . 54
EO254: NEW METHODS IN HIGH DIMENSIONAL DATA ANALYSIS(Room: LSKG003) . . . 55
EO086: REGRESSION AND CLASSIFICATION IN HIGH-DIMENSIONAL SPACES(Room: LSK1010) . . . 55
EO064: HETEROSKEDASTICITY AND AUTOCORRELATION ROBUST INFERENCE(Room: LSK1034) . . . 56
EO070: RECENT ADVANCES IN SPATIAL STATISTICS(Room: LSK1009) . . . 57
EO146: TOPICS IN FINANCIAL AND NONPARAMETRIC ECONOMETRICS(Room: LSK1003) . . . 57
EO038: ADVANCES IN STATISTICAL AND ECONOMETRIC MODELLING OF RISK PROCESSES(Room: LSK1027) . . . 58
EO130: RECURRENT EVENTS(Room: LSK1011) . . . 59
EO278: LEARNING THEORY AND BIG DATA(Room: LSK1007) . . . 59
EO266: ADVANCES IN OPTIMAL PORTFOLIO ALLOCATION AND OPTION PRICING(Room: LSK1033) . . . 60
EO236: THEORY AND NUMERICS IN ESTIMATING STOCHASTIC PROCESSES(Room: LSK1014) . . . 60
EO276: FINANCIAL ECONOMETRICS(Room: LSK1032) . . . 61
Parallel Session K – EcoSta2017 (Saturday 17.06.2017 at 08:30 - 09:50) 62 EO150: ADVANCED GRAPHICAL AND COMPUTATIONAL METHODS(Room: LSKG001) . . . 62
EC286: CONTRIBUTIONS INBAYESIAN ECONOMETRICS(Room: LSK1033) . . . 62
EC292: CONTRIBUTIONS IN FORECASTING(Room: LSKG007) . . . 63
EC294: CONTRIBUTIONS IN STATISTICAL MODELLING(Room: LSK1005) . . . 63
EG129: CONTRIBUTIONS IN SURVIVAL ANALYSIS(Room: LSK1007) . . . 64
EC289: CONTRIBUTIONS IN MULTIVARIATE METHODS(Room: LSK1009) . . . 65
EG061: CONTRIBUTIONS ON TIME SERIES MODELING AND ITS APPLICATIONS(Room: LSK1003) . . . 65
EC284: CONTRIBUTIONS IN TIME SERIES(Room: LSK1001) . . . 66
EG053: CONTRIBUTIONS IN BOOTSTRAP METHODS(Room: LSK1010) . . . 67
EC285: CONTRIBUTIONS IN FINANCIAL ECONOMETRICSI (Room: LSK1034) . . . 67
Parallel Session L – EcoSta2017 (Saturday 17.06.2017 at 10:15 - 11:30) 69 EO296: STATISTICAL MODELS WITH APPLICATIONS(Room: LSK1011) . . . 69
EO222: ADVANCES IN COMPLEX TIME SERIES ANALYSIS AND ITS APPLICATIONS(Room: LSKG007) . . . 69
EO048: CHALLENGES IN FUNCTIONAL DATA ANALYSIS(Room: LSKG001) . . . 70
EO307: RECENT ADVANCES IN CAUSAL INFERENCE METHODS(Room: LSK1010) . . . 70
EO076: CIRCULAR TIME SERIES AND STATISTICAL INFERENCE(Room: LSK1009) . . . 70
EO260: NEW DEVELOPMENTS IN BIOMEDICAL RESEARCHII (Room: LSK1014) . . . 71
EO182: COMPUTATIONAL METHODS IN FINANCIAL STATISTICS(Room: LSK1003) . . . 71
EO030: RECENT ADVANCES IN DYNAMIC PANEL DATA AND FACTOR MODELS(Room: LSK1007) . . . 72
EG239: CONTRIBUTIONS IN FINANCIAL ECONOMETRICSII (Room: LSK1033) . . . 72
EG267: CONTRIBUTIONS IN OPTIMAL PORTFOLIO ALLOCATION AND OPTION PRICING(Room: LSK1034) . . . 72
EC287: CONTRIBUTIONS INBAYESIAN STATISTICS(Room: LSK1005) . . . 73
Parallel Session N – EcoSta2017 (Saturday 17.06.2017 at 14:00 - 15:40) 75
EI004: BAYESIAN NONPARAMETRICS(Room: LSKG001) . . . 75
EO232: MODEL ESTIMATION IN MATHEMATICAL FINANCE(Room: LSK1026) . . . 75
EO088: QUANTILE REGRESSION IN HIGH DIMENSIONS(Room: LSK1001) . . . 75
EO078: MODELING AND TESTING PROBLEMS WITH COMPLEX HIGH-DIMENSIONAL DATA(Room: LSK1005) . . . 76
EO170: RECENT ADVANCES IN TIME SERIES ANALYSIS(Room: LSK1032) . . . 77
EO102: NEW ADVANCES IN STATISTICAL MODELING,COMPUTATION AND APPLICATIONS(Room: LSK1009) . . . 77
EO056: PERFORMANCE ANALYSIS(Room: LSK1003) . . . 78
EO311: RECENT ADVANCES IN JOINT MODELING(Room: LSK1034) . . . 78
EO028: FORECASTING ECONOMIC AND FINANCIAL TIME SERIES(Room: LSKG007) . . . 79
EO194: NONPARAMETRIC METHODS FOR VARIABILITY ESTIMATION(Room: LSK1011) . . . 80
EO090: RECENT DEVELOPMENTS IN TIME SERIES ANALYSIS AND RELATED TOPICS(Room: LSK1027) . . . 80
EO172: STATISTICAL INFERENCE AND THEIR APPLICATIONS TO COMPLEX PROBLEMS(Room: LSK1007) . . . 81
EO242: BAYESIAN MODELING FOR SPATIOTEMPORAL PHENOMENA(Room: LSK1033) . . . 81
EO106: LARGE-SCALE REGRESSION METHODS AND ALGORITHMS (Room: LSK1010) . . . 82
EO198: RECENT DEVELOPMENT IN STATISTICAL ANALYSIS OF FUNCTIONAL AND IMAGE DATA(Room: LSK1014) . . . 82
Parallel Session O – EcoSta2017 (Saturday 17.06.2017 at 16:10 - 17:50) 84 EO082: ADVANCES IN HIGH-DIMENSIONAL DATA ANALYSIS(Room: LSK1007) . . . 84
EO120: NONLINEAR TIME SERIES(Room: LSK1033) . . . 84
EO040: ENDOGENEITY AND NONPARAMETRICS IN MODELS OF PRODUCTION(Room: LSK1003) . . . 85
EO202: NONPARAMETRIC AND SEMI PARAMETRIC STATISTICS AND THEIR APPLICATIONS(Room: LSK1011) . . . 85
EO188: APPLICATIONS AND EMPIRICAL RESEARCH IN ECONOMICS AND FINANCE(Room: LSK1001) . . . 86
EO098: RECENT DEVELOPMENTS IN ECOLOGICAL STATISTICS(Room: LSK1009) . . . 87
EO142: RECENT ADVANCES INBAYESIAN COMPUTATION(Room: LSK1010) . . . 87
EO210: NEW DEVELOPMENTS IN FUSION LEARNING AND STATISTICAL INFERENCES(Room: LSKG001) . . . 88
EO186: SUFFICIENT DIMENSION REDUCTION IN SURVIVAL ANALYSIS(Room: LSK1005) . . . 89
EO068: ADVANCES IN VOLATILITY MODELLING AND FORECASTING(Room: LSKG007) . . . 89
EO268: SPATIAL ECONOMETRICS(Room: LSK1027) . . . 90
EO204: FUNCTIONAL DATA ANALYSIS AND ITS APPLICATIONS(Room: LSK1014) . . . 90
EO316: FINANCIAL INTEGRATION AND CRISIS TRANSMISSION(Room: LSK1034) . . . 91
EO160: NEW DEVELOPMENTS IN FINANCIAL TIME SERIES(Room: LSK1032) . . . 92 93
Thursday 15.06.2017 09:00 - 09:50 Room: LTA - Citi Lecture Theater Chair: Ana Colubi Keynote talk 1 A statistical tale of subgroup analysis for managerial decision making
Speaker: Xuming He, University of Michigan, United States
Clinical trials are often used for the assessment of therapeutic benefits. Subgroup analysis is routinely used in the clinical studies to understand heterogeneity of treatment effects in subgroups of patients. When used appropriately, subgroup analysis helps personalized treatments of patients and better designs of new studies. When used indiscriminately, subgroup analysis leads to costly false discoveries. We will start with a partial review of subgroup identification and confirmation methods, and examine the benefits and pitfalls of subgroup analysis associated with decision making in the pharmaceutical industry. We will discuss some important questions of how the risk for subgroup pursuit needs to be quantified to help managerial decision making. In this process, we identify several interesting statistical questions.
Thursday 15.06.2017 17:40 - 18:30 Room: LTA - Citi Lecture Theater Chair: Alan Wan Keynote talk 2 Quantile spectral analysis for locally stationary time series
Speaker: Marc Hallin, Universite Libre de Bruxelles, Belgium Stefan Birr, Holger Dette, Stanislav Volgushev Classical spectral methods are subject to two fundamental limitations: they only can account for covariance-related serial dependencies, and they require second-order stationarity. Much attention has been devoted lately to quantile-based spectral methods that go beyond covariance-based serial dependence features. At the same time, covariance-based methods relaxing stationarity into much weaker local stationarity conditions have been developed for a variety of time-series models. We are combining those two approaches by proposing quantile-based spectral methods for locally stationary processes. We therefore introduce a time-varying version of the copula spectra that have been recently proposed in the literature, along with a suitable local lag-window estimator. We propose a new definition of local strict stationarity that allows us to handle completely general non-linear processes without any moment assumptions, thus accommodating our quantile-based concepts and methods. We establish a central limit theorem for the new estimators, and illustrate the power of the proposed methodology by means of a simulation study. Moreover, in two empirical studies, we demonstrate that the new approach detects important variations in serial dependence structures both across time and across quantiles. Such variations remain completely undetected, and are actually undetectable, via classical covariance-based spectral methods.
Friday 16.06.2017 14:00 - 14:50 Room: LTA - Citi Lecture Theater Chair: Erricos Kontoghiorghes Keynote talk 3 Robust normal mixtures for financial portfolio allocation
Speaker: Marc Paolella, University of Zurich, Switzerland
A new approach for multivariate modeling and prediction of asset returns is proposed. It is based on a two-component normal mixture, estimated using a fast new variation of the minimum covariance determinant (MCD) method made suitable for time series. It outperforms the (shrinkage-augmented) MLE in terms of out-of-sample density forecasts and portfolio performance. In addition to the usual stylized facts of skewness and leptokurtosis, the model also accommodates leverage and contagion effects, but is i.i.d., and thus does not embody, for example, a GARCH-type structure. Owing to analytic tractability of the moments and the expected shortfall, portfolio optimization is straightforward, and, for daily equity returns data, is shown to substantially outperform the equally weighted and classical long-only Markowitz framework, as well as DCC-GARCH (despite not using any kind of GARCH-type filter).
Saturday 17.06.2017 11:40 - 12:30 Room: LTA - Citi Lecture Theater Chair: Mike So Keynote talk 4 Recent developments in Bayesian inference for time series
Speaker: Michael Pitt, Kings College London, United Kingdom
The pseudo-marginal algorithm is a variant of the Metropolis Hastings scheme which samples asymptotically from a target probability density when we are only able to estimate unbiasedly an unnormalized version of it. It has found numerous applications in statistics and econometrics as there are many scenarios where the likelihood function is intractable but can be estimated unbiasedly using Monte Carlo samples. Several recent contributions will be discussed which optimise the trade off between computational complexity and statistical efficiency. A modification of the pseudo-marginal algorithm, termed the correlated pseudo-marginal algorithm, is introduced. Guidelines are provided for the optimal settings of this algorithm. The computational gains of the new algorithm are demonstrated by examining large time series, including the estimation of continuous time volatility models.
Thursday 15.06.2017
10:25 - 12:30
Parallel Session B – EcoSta2017
EO116 Room LSK1005 NEW DEVELOPMENT IN ANALYZING LARGE COMPLEX DATA Chair: Xuming He
EO0152: Lack-of-fit tests for quantile regression models
Presenter: Xingdong Feng, Shanghai University of Finance and Economics, China
The aim is to novelly transforms lack-of-fit tests for parametric quantile regression models into checking the equality of two conditional distributions of covariates. We then can borrow these successful test statistics from the rich literature of two-sample problems, and this gives us much flexibility in constructing a reliable test according our experiences on covariates. As an illustration, three two-sample test statistics are considered. The first one will lead to an already known test in this context when a Cramer-von Mises test statistic is employed. The second one is a practical two-sample test statistic especially for multivariate distributions, and the resulting test is still applicable in real applications when the number of covariates is moderate or even large. In the last case, we provide a lack-of-fit test based on two-sample tests on moments for the high dimensional data. Its usefulness is demonstrated by simulation experiments and a real example.
EO0591: Test for high dimensional regression coefficients using refitted cross-validation variance estimation Presenter: Wei Zhong, Xiamen University, China
Testing a hypothesis for high dimensional regression coefficients is of fundamental importance in the statistical theory and applications. The aim is to develop a new test for coefficients in high dimensional linear regression models based on an estimated U-statistics of order two. With the aid of martingale central limit theorem, we prove that the asymptotic null distributions of the proposed test are normal under two different distribution assumptions. The idea of refitted cross-validation (RCV) approach is utilized to reduce the bias of the sample variance in the estimation of the test statistic. We assess the finite-sample performance of the proposed test by examining its size and power via Monte Carlo simulations which show that the new test based on the RCV estimator of the variance achieves higher powers, especially for the sparse cases. We also illustrate the application of the proposed test by an empirical analysis of a microarray data set on Yorkshire gilts.
EO0651: Model-free variable selection
Presenter: Junhui Wang, City University of Hong Kong, Hong Kong Co-authors: Xin He, Shaogao Lyu
Variable selection is central to sparse modeling, and many methods have been proposed under various model assumptions. We will present a model-free variable selection method that allows for general variable effects. As opposed to most existing methods based on an explicit functional relationship, the proposed method attempts to identify non-informative variables that are conditional independent with the response by simultane-ously examining the sparsity in multiple conditional quantile functions. It does not require specification of the underlying model for the response, which is appealing in sparse modeling with a relatively large number of variables. The proposed method is implemented via an efficient computing algorithm, which couples the majorize-minimization algorithm and the proximal gradient descent algorithm. The effectiveness of the proposed method is also supported by a variety of simulated and real-life examples. Its asymptotic estimation and variable selection consistencies are established without explicit model assumption.
EO0829: High dimensional variable selection for longitudinal data with covariate measurement error and dropout Presenter: Yang Bai, Shanghai University of Finance and Economics, China
A novel approach is developed for high-dimensional variable selection in longitudinal data with covariate measurement error and dropout. Specifi-cally, new penalized estimating equations are proposed for variable selection, while measurement error and dropout are dealt with simultaneously. A modified coordinate descent algorithm for solving this specific variable selection problem is developed. The asymptotic properties of the pro-posed estimators are established in high dimensional settings. We demonstrate that the propro-posed approach is consistent in variable selection and achieves the oracle property under regularity conditions. Finite sample performance of the proposed method is evaluated via both extensive Monte Carlo simulation and a real data analysis in systematic lupus erythematosus.
EO0249: Statistical inference for the single index hazards model Presenter: Sheng Xu, The Hong Kong Polytechnic University, Hong Kong Co-authors: Jicai Liu, Catherine Liu, Xihong Lin
A class of consistent estimators is developed for the parameter vector for the single-index hazards model, where the model allows nonparametric modeling of covariate effects in a parsimonious way through a single index and circumvent the curse of dimensionality. A class of consistent profile likelihood estimators is proposed by estimating the nonparametric part via local linear smoothing tool, and hence the estimator of the parametric component is shown to be asymptotically normal and even achieves the semiparametric efficiency bound. Motivated by the semiparametric efficient score, we propose two classes of estimation equations such that their roots enjoys doubly robust properties. We present some algorithm procedures for validity of model checking. Extensive numerical analysis are implemented so as to assess the finite-sample properties of our proposed estimators compared with other existed methods. We also demonstrate our method by several real-world data application.
EO018 Room LSK1007 STATISTICAL MODELLING FOR NETWORK DATA Chair: Yang Feng
EO0163: Graph-based change-point analysis for object data Presenter: Hao Chen, University of California at Davis, United States
After observing snapshots of a network, the aim is to check whether there has been a change in dynamics. We develop a general nonparametric framework for change-point detection that relies on a distance metric on the sample space of observations. This new approach, which relies on graph-based tests, can be applied to high dimensional data, as well as data from non-Euclidean sample spaces, such as network data. An analytic approximation for the false positive error probability is derived and shown to be reasonably accurate by simulation. We illustrate the method through the analysis of a phone-call network from the MIT Reality Mining.
EO0203: A semidefinite program for the stochastic blockmodel Presenter: David Choi, Carnegie Mellon University, United States
Semidefinite programs have recently been developed for the problem of community detection, which may be viewed as a special case of the stochastic blockmodel. We develop a semidefinite program that can be tailored to other instances of the blockmodel, such as non-assortative networks and overlapping communities. We establish label recovery in sparse settings, with conditions that are analogous to recent results for community detection. In settings where the data is not generated by a blockmodel, we give an oracle inequality that bounds excess risk relative to the best blockmodel approximation. Simulations are presented for community detection, for overlapping communities, and for latent space models.
EO0519: Estimating network memberships by simplex vertices hunting Presenter: Tracy Ke, University of Chicago, United States
Consider an undirected mixed-membership network with n nodes and K communities. For each node i, we model the membership by a Probability Mass Function (PMF)πi= (πi(1),πi(2), . . .,πi(K))0, whereπi(k) is the probability that node i belongs to community k. We call node i ‘pure’ if πiis degenerate and ‘mixed’ otherwise. The primary interest is to estimateπi, 1≤ i ≤ n. We propose Mixed-SCORE as a new spectral method for mixed membership estimation. At the heart of Mixed-SCORE is a (tall by very skinny) matrix of entry-wise ratios, formed by dividing by the first few eigenvectors of the network adjacency matrix over the leading eigenvector of the same matrix in an entry-wise fashion. The main surprise is that the rows of the entry-wise ratio matrix form a cloud of points in a low-dimensional space with the silhouette of a simplex, which simplex carries all information we need for estimating the memberships. We apply Mixed SCORE to four network data sets (a coauthorship and a citee network for statisticians, a political book network, and a football network) and obtain meaningful results. We propose a Degree-Corrected Mixed Membership (DCMM) model, and use it to solidify our discoveries with delicate spectral analysis and Random Matrix Theory.
EO0628: Random walk models of network formation Presenter: Peter Orbanz, Columbia University, United States
A class of network models are described that insert edges by connecting the starting and terminal vertices of a random walk on the network graph. Within the taxonomy of statistical network models, this class is distinguished by permitting the location of a new edge to explicitly depend on the structure of the graph, but being nonetheless statistically and computationally tractable. In the limit of infinite walk length, the model converges to an extension of the preferential attachment model. Theoretical properties will be discussed, such as power laws, and show that inference of model parameters is possible from a single graph generated by the model.
EO0192: Multivariate spatial autoregression for large scale social network Presenter: Xuening Zhu, Peking University, China
The rapid growth of social network platforms generates a large amount of social network data. As a result, multivariate responses and the cor-responding predictors can be collected for social network users. To statistically model such type of data, the multivariate spatial autoregression (MSAR) model is proposed and studied. To estimate the model, the maximum likelihood estimator (MLE) is obtained under certain technical conditions. However, it is found that the computational cost of MLE is expensive. In order to fix this problem, a least squares type estimator is developed. The corresponding asymptotic properties are investigated. To gauge the finite sample performance of the proposed estimators, a number of numerical studies are conducted. Lastly, a Sina Weibo dataset is analyzed for illustration purpose.
EO136 Room LSK1011 STATISTICAL METHODS FOR FUNCTIONAL DATA Chair: Yuhang Xu
EO0189: Nested hierarchical functional data modeling and inference for evaluating lunar effects on root gravitropism Presenter: Yuhang Xu, University of Nebraska - Lincoln, United States
Co-authors: Dan Nettleton
In a plant science study, the process of roots bending in response to gravity is recorded. The data are collected from seeds representing a large variety of genotypes and have a three-level nested hierarchical structure. We allow the mean function of the bending rate to depend on the lunar day and model the variation between genotypes, groups of seeds imaged together, and individual seeds by hierarchical functional random effects. We estimate the covariance functions by a fast penalized tensor product spline approach, perform multi-level functional PCA using the BLUP of the principal component scores, and improve the efficiency of mean estimation by iterative decorrelation. We choose the number of principal components using a conditional AIC and test the lunar day effect using generalized likelihood ratio test statistics based on the marginal and conditional likelihoods. We propose a permutation procedure to evaluate the null distribution of the test statistics. Simulation studies show that our model selection criterion selects the correct number of principal components with high frequency, and the likelihood-based tests based on functional PCA have higher power than a test based on working independence. Our data analysis suggests that the root bending behavior may be associated with moon phases.
EO0234: Function-on-function regression for highly densely observed spiky functional data Presenter: Ruiyan Luo, Georgia State University, United States
Modern techniques allow data to be recorded with high sample rate, which leads to highly densely observed spiky curves. For example, the mass spectrometry data contains a number of narrow and high peaks which are interests of scientists, and the EEG curves exhibit high local variations over the whole time interval. The existing methods for function-on-function regression assume that the coefficient functions are smooth (usually they are assumed to belong to the Sobolev space), and impose smoothness regularities in various ways. However, the smoothness assumption makes it difficult to model the associations between high local variations in response curve and predictor curves. We model the coefficient functions in a more general family of function spaces, where various levels of local variations are possible. We propose a new regularization method to replace the traditional smoothing method in functional data analysis, and apply it to our recently developed signal compression method in function-on-function regression. In addition to capturing the association between high local variations, such as rapid peaks, this method has good prediction performance and can handle multiple functional predictors with thousands of observation time points.
EO0235: Nonparametric operator-regularized covariance function estimation Presenter: Raymond Wong, Iowa State University, United States
Co-authors: Xiaoke Zhang
A class of nonparametric covariance function estimators is developed by utilizing spectral regularization of an operator, which is associated with a typically infinite dimensional reproducing kernel Hilbert space. By construction, these estimators are positive semi-definite and hence valid covariance functions. A related representer theorem is established to provide a finite dimensional representation of such estimators. In order to achieve low-rank estimations, trace-norm regularization is studied in detail. A specific computational algorithm is developed and this estimator is shown to enjoy excellent rates of convergence under either fixed or random designs. The empirical performance of the proposed trace-norm-regularized estimator is demonstrated in a simulation study, while its practical utility is illustrated in an analysis of a traffic data set.
EO0348: Hypothesis testing in functional linear models
Presenter: Yu-Ru Su, Fred Hutchinson Cancer Research Center, United States Co-authors: Chong-Zhi Di, Li Hsu
Functional data arise frequently in biomedical studies, where it is often of interest to investigate the association between functional predictors and a scalar response variable. While functional linear models (FLM) are widely used to address these questions, hypothesis testing for the functional association in the FLM framework remains challenging. A popular approach to testing the functional effects is through dimension reduction by functional principal component (PC) analysis. However, its power performance depends on the choice of the number of PCs, and is not systematically studied. We first present the power performance of the Wald-type test with varying thresholds in selecting the number of PCs for the functional covariates, and show that the power is sensitive to the choice of thresholds. To circumvent the issue, we propose a new method of ordering and selecting principal components to construct test statistics. The proposed method takes into account both the association
with the response and the variation along each eigenfunction. We establish its theoretical properties and assess the finite sample properties through simulations. Our simulation results show that the proposed test is more robust against the choice of threshold while being as powerful as, and often more powerful than, the existing method. We then apply the proposed method to the cerebral white matter tracts data obtained from a diffusion tensor imaging tractography study.
EO0838: Component selection and estimation for functional additive models Presenter: Hao Zhang, University of Arizona, United States
Functional additive model provides a flexible yet simple framework for regressions involving functional predictors. The utilization of data-driven basis in an additive rather than linear structure naturally extends the classical functional linear model. However, the critical issue of selecting nonlinear additive components has been less studied. We propose a new regularization framework for joint component selection and estimation in the context of the Reproducing Kernel Hilbert Space. The proposed approach takes advantage of the functional principal components which greatly facilitates the implementation and the theoretical analysis. The selection and estimation are achieved by penalized least squares using a penalty which encourages the sparse structure of the additive components. Theoretical properties, such as the existence and the rate of convergence are investigated. The empirical performance is demonstrated through simulation studies and a real data application.
EO108 Room LSK1001 MODEL AVERAGING,SELECTION AND SHRINKAGE Chair: Alan Wan
EO0205: PMSE performance of two different types of preliminary test estimators under a multivariate t error Presenter: Haifeng Xu, Xiamen University, China
Co-authors: Kazuhiro Ohtani
Assuming that the error terms follow a multivariate t distribution, the exact formula is derived for the predictive mean squared error (PMSE) of two different types of preliminary test estimators: (1) a homogeneous pre-test (HO-PT) estimator whose components are the adjusted minimum mean squared error (AMMSE) estimator and the minimum mean squared error (MMSE) estimator; (2) a heterogeneous pre-test (HE-PT) estimator whose components are the AMMSE estimator and the Stein-rule (SR) estimator. It is shown analytically that the HE-PT estimator dominates the SR estimator if a critical value of the pre-test is chosen appropriately. Also, we compare the PMSE of the HO-PT, HE-PT, MMSE, AMMSE, SR and PSR estimators by numerical evaluations. Our results show that 1. the HO-PT and HE-PT estimators dominate the OLS estimator for all combinations when the degrees of freedom is not more than 5; 2. if the number of independent variables is 3, and the critical value of the pre-test is chosen appropriately, then the HE-PT estimator dominates the PSR estimator even when error terms follow a multivariate t distribution. EO0391: Dominance of the positive-part shrinkage estimator when each individual regression coefficient is estimated
Presenter: Akio Namba, Kobe University, Japan Co-authors: Haifeng Xu
Assuming that there exist omitted variables in the specified model, the aim is to analytically derive the exact formula for the mean squared error (MSE) of a general family of shrinkage estimators for each individual regression coefficient. It is shown analytically that when our concern is to estimate each individual regression coefficient, the positive-part shrinkage estimators dominate the shrinkage estimators under some conditions even when the relevant regressors are omitted. Also, by numerical evaluations, we showed the effects of our theorem for several specific cases. It is shown that the positive-part shrinkage estimators dominate shrinkage estimators for large region of parameter space even when there exist omitted variables in the specified model.
EO0221: Conditionally optimal weights: Optimal combination with predictable errors Presenter: Andrey Vasnev, University of Sydney, Australia
A classical unconditional framework is extended and conditionally optimal weights are constructed that it can be used to combine individual forecasts. Often there is an information set which is available when the combination is constructed. The previous forecast errors can be included in this information set, but it could also contain other variables. If conditionally on this information the forecast errors are predictable, then the conditional mean squared error (MSE) of the combination needs to be minimized rather than unconditional variance used in the classical framework. We prove that the new conditionally optimal weights produce combinations with smaller expected MSE than the classical unconditionally optimal weights. Our empirical study of the European Central Bank Survey of Professional Forecasters confirms theoretical findings and shows that the new weights outperform a wide range of other methods.
EO0219: Model averaging in a multiplicative heteroscedastic model Presenter: Alan Wan, City University of Hong Kong, Hong Kong
In recent years, the literature of frequentist model averaging in econometrics has grown significantly. Models with different mean structures have been considered, but variance considerations have been left out. We consider a regression model with multiplicative heteroscedasticity and develop a model averaging method that combines maximum likelihood estimators of unknown parameters in both the mean and variance functions of the model. Our weight choice criterion is based on a minimisation of a plug-in estimator of the model average estimator’s squared prediction risk. We prove that the new estimator possesses an asymptotic optimality property. Our investigation of finite-sample performance by simulations demonstrates that the new estimator frequently exhibits very favourable properties compared to some existing heteroscedasticity-robust model average estimators. The model averaging method hedges against the selection of very bad models and serves as a remedy to variance function mis-specification, which often discourages practitioners from modeling heteroscedasticity altogether. The proposed model average estimator is applied to the analysis of two data sets on housing and economic growth.
EO0225: Focused information criterion and model averaging for large panels with a multifactor error structure Presenter: Chu-An Liu, Academia Sinica, Taiwan
Co-authors: Chang-Ching Lin, Shou-Yung Yin
Model selection and model averaging is considered for panel data models with a multifactor error structure. We investigate the limiting distribution of the common correlated effects estimator in a local asymptotic framework and show that the trade-off between bias and variance remains in the asymptotic theory. We then propose a focused information criterion and a plug-in averaging estimator for large heterogeneous panels and examine their theoretical properties. The novel feature of the proposed method is that it aims to minimize the sample analogue of the asymptotic mean squared error and can be applied to cases irrespective of whether the rank condition holds or not. Monte Carlo simulations show that both proposed selection and averaging methods generally achieve lower expected squared error than other methods. The proposed methods are applied to analyze the consumer response to gasoline taxes.
EO112 Room LSK1034 NEW DEVELOPMENTS IN FINANCIAL ECONOMETRICS Chair: Daniel Preve
EO0242: Asymptotic properties of maximum likelihood estimators of a multiplicative time-varying correlation GARCH model Presenter: Timo Terasvirta, Aarhus University, Denmark
A new multivariate volatility model that belongs to the family of conditional correlation GARCH models is introduced. The GARCH equations of this model contain a multiplicative deterministic component to describe long-run movements in volatility and, in addition, the correlations are deterministically time-varying. Parameters of the model are estimated jointly using maximum likelihood. Consistency and asymptotic normality of maximum likelihood estimators is proved. Numerical aspects of the estimation algorithm are discussed. A bivariate empirical example is provided. EO0449: High dimensional minimum variance portfolio estimation with high-frequency data
Presenter: Yingying Li, Hong Kong University of Science and Technology, Hong Kong Co-authors: Tony Cai, Jianchang Hu, Xinghua Zheng
The estimation of high dimensional minimum variance portfolio (MVP) with high frequency financial data is considered. High frequency returns can exhibit heteroskedasticity and possiblybe contaminated by microstructure noise. Under some sparsity assumptions on the precision matrix, we propose an estimator of MVP, which asymptotically achieves the minimum variance. Simulation and empirical studies demonstrate that our proposed portfolio performs favorably.
EO0525: A mixture autoregressive model based on Student’s t-distribution Presenter: Mika Meitz, University of Helsinki, Finland
Co-authors: Daniel Preve, Pentti Saikkonen
A new mixture autoregressive model is proposed based on Student’s t-distribution. A key feature of our model is that the conditional t-distributions of the component models are based on autoregressions that have t-distributions as their stationary distributions. That autoregressions with such stationary distributions exist is not immediate. Our formulation implies that the conditional variance of each mixture component is not constant but of (nonlinear) ARCH type. Compared to previous mixture autoregressive models our model may therefore be useful in applications where the data exhibits rather strong conditional heteroskedasticity. Our formulation also has the theoretical advantage that conditions for stationarity and ergodicity are always met and these properties are much more straightforward to establish than is common in nonlinear autoregressive models. An empirical example employing a realized kernel series based on S&P 500 data shows that the proposed model performs well in volatility forecasting. EO0212: The risk return relationship: Evidence from index return and realised variance series
Presenter: Minxian Yang, The University of New South Wales, Australia
The risk return relationship is analysed in bivariate models for return and realised variance (RV) series. Based on daily time series from 21 international market indices for more than 13 years (January 2000 to February 2013), the empirical findings support the arguments of risk return tradeoff, volatility feedback and statistical balance. It is argued that the empirical risk return relationship is primarily shaped by two important data features: the negative contemporaneous correlation between the return and RV, and the difference in the autocorrelation structures of the return and RV. The findings do not support the risk premium effect of the shock to volatility as documented by recent studies that do not take into account of the contemporaneous correlation.
EO0508: New distribution theory for the estimation of structural break point in mean Presenter: Xiaohu Wang, The Chinese University of Hong Kong, Hong Kong
Co-authors: Liang Jiang, Jun Yu
Based on the Girsanov theorem, the exact distribution of the maximum likelihood estimator of structural break point in a continuous time model is obtained. The exact distribution is asymmetric and tri-modal, indicating that the estimator is biased. These two properties are also found in the finite sample distribution of the least squares (LS) estimator of structural break point in the discrete time model, suggesting the classical long-span asymptotic theory is inadequate. A continuous time approximation to the discrete time model is built, and an in-fill asymptotic theory for the LS estimator is developed. The in-fill asymptotic distribution is asymmetric and tri-modal and delivers good approximations to the finite sample distribution. To reduce the bias in the estimation of both the continuous time and the discrete time models, a simulation-based method based on the indirect estimation (IE) approach is proposed. Monte Carlo studies show that IE achieves substantial bias reductions.
EO010 Room LSK1027 MODELLING FINANCIAL AND INSURANCE RISKS Chair: Tak Kuen Siu
EO0277: On modeling credit defaults: A probabilistic Boolean network approach Presenter: Wai-Ki Ching, The University of Hong Kong, Hong Kong
One of the central issues in credit risk measurement and management is modeling and predicting correlated defaults. A novel model is introduced to investigate the relationship between correlated defaults of different industrial sectors and business cycles as well as the impacts of business cycles on modeling and predicting correlated defaults using the Probabilistic Boolean Network (PBN). The key idea of the PBN is to decompose a transition probability matrix describing correlated defaults of different sectors into several BN matrices which contain information about business cycles. An efficient estimation method based on entropy approach is used to estimate the model parameters. Using real default data, we build a PBN for explaining the default structure and make reasonably good prediction of joint defaults in different sectors.
EO0502: Longevity-product valuation under correlated financial and mortality risks Presenter: Rogemar Mamon, University of Western Ontario, Canada
The pricing of annuity and guaranteed annuity option primarily depends on mortality and interest risk factors. Independence between these two risk factors greatly facilitates the valuation but such an assumption is not always realistic. We propose a pricing framework where the dependence between interest and mortality rates is modelled explicitly. We employ the change of measure technique in conjunction with the comonotonicity theoretic approach to approximate annuity and guaranteed annuity option prices. Our practical method provides accurate pricing values and is efficient as it circumvents the simulation-within-simulation problem.
EO0453: Rating of financial products by implied risk aversion and optimized expected utility risk measures Presenter: Joern Sass, University of Kaiserslautern, Germany
Co-authors: Holger Fink, Sebastian Geissel, Frank Thomas Seifried
An optimal expected utility risk measure (OEU) is introduced which is generated by a utility function via an associated optimal investment problem: A financial product is evaluated by finding the capital to be borrowed and added to the position in order to maximize the discounted certainty equivalent of the future payoff. Properties of OEU are derived and put in relation to alternative risk measures. For constant relative risk aversion and for proper discounting, OEU is non-trivial and coherent. OEU reacts in a more sensitive way to slight changes of the probability of a financial loss than (average) value at risk. This motivates to use implied risk aversion based on OEU as a coherent rating methodology for structured financial products. This takes into account both upside potential and downside risks and is easily interpreted in terms of an individual investor’s
