Calendar

Feb
12
Wed
Beyond Measurement Artifacts: Integrating Measurement Equivalence with Theory Development in Cross Cultural Research @ York University Schulich School of Business, N106
Feb 12 @ 11:30 – 13:00
The Organization Studies area at Schulich invites you to attend a seminar with Professor Gordon Cheung (The Chinese University of Hong Kong). Prof. Cheung is an outstanding and innovative researcher with expertise in research methods and structural equation modelling, as well as international and cross-cultural research.
Prof. Cheung is currently professor at the Department of Management, The Chinese University of Hong Kong.  He is a dedicated researcher with expertise in research methods and structural equation modeling. He has published more than 20 articles in research methodologies, which have been cited about 4,000 times. He has twice received the Sage Best Paper Award from the Research Methods Division of the Academy of Management (2000 and 2009) and in 2008 the Best Published Paper Award in Organizational Research Methods. Prof. Cheung served as the Division Chair of the Research Methods Division of the Academy of Management in 2006/07. Prof. Cheung’s research interest in measurement equivalence/invariance (ME/I) started more than 15 years ago and he has published over 10 papers in this area. His paper “Testing Factorial Invariance Across Groups: A Reconceptualization and Proposed New Method” published at Journal of Management in 1999 and the paper “Assessing Extreme and Acquiescence Response Sets in Cross-Cultural Research Using Structural Equations Modeling” published at Journal of Cross-Cultural Psychology in 2000 define the way on how ME/I should be examined. The paper “Evaluating Goodness-of-Fit Indices for Testing Measurement Invariance” published in 2002 at Structural Equation Modeling Journal, which defines the standard on how nested models should be compared, has received more than 2,000 citations. The paper “Testing Equivalence in the Structure, Means, and Variances of Higher-Order Constructs with Structural Equation Modeling” published in 2008 at Organizational Research Methods received the 2008 Best Paper Published in Organizational Research Methods Award.Structure, Means, and Variances of Higher-Order Constructs with Structural Equation Modeling” published in 2008 at Organizational Research Methods received the 2008 Best Paper Published in Organizational Research Methods Award. 

Feb
13
Thu
Computational Foundations of Bayesian Inference and Probabilistic Programming @ Sidney Smith Hall, Room 1074
Feb 13 @ 15:30 – 16:45

Thursday February 13, 2014

at 3:30pm

Sidney Smith Hall, Room 1074

**Refreshments will be served at 3:15pm

Computational Foundations of Bayesian Inference and Probabilistic Programming

Dr. Daniel Roy, University of Cambridge

The complexity, scale, and variety of data sets we now have access to have grown enormously, and present exciting opportunities for new applications.  Just as high-level programming languages and compilers empowered experts to solve computational problems more quickly, and made it possible for non-experts to solve them at all, a number of high-level probabilistic programming languages with computationally universal inference engines have been developed with the potential to similarly transform the practice of Bayesian statistics.  These systems provide formal languages for specifying probabilistic models compositionally, and general algorithms for turning these specifications into efficient algorithms for inference.

In this talk, I will address three key questions at the theoretical and algorithmic foundations of probabilistic programming—and probabilistic modeling more generally—that can be answered using tools from probability theory, computability and complexity theory, and nonparametric Bayesian statistics.  Which Bayesian inference problems can be automated, and which cannot?  Can probabilistic programming languages represent the stochastic processes at the core of state-of-the-art nonparametric Bayesian models?  And if not, can we construct useful approximations?  I’ll close by relating these questions to other challenges and opportunities ahead at the intersections of computer science, statistics, and probability.

 

 

http://www.utstat.toronto.edu/wordpress/?page_id=18

 

Feb
24
Mon
Integrating Ratings of Child Psychopathology across Multiple Informants @ Norm Endler Room (BSB 164)
Feb 24 @ 22:15 – 23:15
Quantitative Methods Forum @ Norm Endler Room (BSB 164)

Feb 24 @ 10:15 AM – 11:15 AM

Speaker: Dr. Andrea Howard, Carleton University
Department of Psychology

Title: Integrating Ratings of Child Psychopathology across Multiple Informants

Abstract: One of the most significant challenges facing researchers and practitioners who assess child psychopathology is how to integrate information about a child’s symptoms from multiple sources when those sources provide discrepant ratings (De Los Reyes & Kazdin, 2004). It is common to obtain ratings for a single target child from informants such as parents, teachers, and peers, but it is less clear how to combine the information provided by multiple informants to derive an integrated measure of the psychopathology trait of interest that is not confounded with informants’ unique perspectives. A new approach to this problem stipulates a  trifactor measurement model to analytically disaggregate informants’ unique perspectives of children’s symptoms from a cross-informant consensus rating of their true symptoms (Bauer, Howard et al., 2013). Preliminary results from a new study expand the trifactor model to a three-informant, multi-trait assessment of inattention and hyperactivity/impulsivity symptoms using data drawn from baseline assessments of children enrolled in a randomized controlled trial study of treatments for Attention-Deficit/Hyperactivity Disorder (ADHD). 

Suggested Readings:
Bauer, D. J., Howard, A. L., Baldasaro, R. E., Curran, P. J., Hussong, A. M., Chassin, L., & Zucker, R. A. (2013). A trifactor model for integrating ratings across multiple informants . Psychological methods, 18(4), 475-493.

Feb
27
Thu
Data mining with R: Let R ‘rattle’ you @ Ted Rogers School of Management, TRS 2-166 (8th floor)
Feb 27 @ 14:00 – 16:00

Title:  Data mining with R: Let R ‘rattle’ you

 

Description: 

 

This hands-on workshop will provide training in the rattle data mining package for R. rattle is a graphical user interface to transform, visualize and analyze data.

 

For hands-on exercises, please bring a laptop installed with R and rattle.

 

**R is available at http://probability.ca/cran/. Instructions for the rattle installation are available at http://rattle.togaware.com/rattle-install-mswindows.html or http://rattle.togaware.com/rattle-install-mac.html

 

Instructor:

 

Murtaza Haider

Associate Dean of Research & Graduate Programs

Ted Rogers School of Management

Ryerson University

 

Date and Time:  Thursday, February 27, 2014 (2:00 pm — 4:00 pm)

 

Location: 

 

Ted Rogers School of Management

55 Dundas Street West

Room TRS 2-166 (8th floor)

 

RSVP: Nik Ashton (nashton@ryerson.ca)

 

Feb
28
Fri
On a multiple-shock dependence structure @ North Ross 638
Feb 28 @ 10:30 – 11:30
 Su Jianxi supervised by Prof. Edward Furman will be giving a talk in the York University Statistics seminar series.
The title of the talk is: On a multiple-shock dependence structure. The seminar will be given on Friday Feb 28, 10:30-11:30 at N638Ross.
Mar
6
Thu
**CANCELLED** Backward Simulation of Poisson Processes @ Sidney Smith Room 1074
Mar 6 @ 15:30 – 16:30

SEMINAR

Thursday March 6, 2014 at 3:30pm
Sidney Smith Hall, Room 1074

**Refreshments will be served at 3:15pm**

Dr. Alexander Kreinin, IBM
Head of Quantitative Research, RFE Risk Analytics, Business Analytics

Backward Simulation of Poisson Processes

Multivariate Poisson processes have many applications in financial modeling. In particular, in the area of Operational Risk they are used for description and simulation of the frequencies of operational events. Practitioners often model the operational events independently despite observed correlations between the components. In this talk we discuss simulation of the multivariate Poisson model based on the “Poisson Bridge” idea, extreme correlations and their dependence on the intensities of the processes.

Mar
10
Mon
Evaluation of Intimate Partner Risk Assessment Inventories @ Norm Endler Room (BSB 164)
Mar 10 @ 10:15 – 11:30

Quantitative Methods Forum @ Norm Endler Room (BSB 164)

Mar 10 @ 10:15 AM – 11:15 AM

Speaker: Carrie Smith, York University
Department of Psychology

Title: Evaluation of Intimate Partner Risk Assessment Inventories

Abstract: In the interest of improving transparency, replicability, and validity, many jurisdictions and agencies are now favouring the use of empirically validated measures in violence risk assessment. To date, dozens of risk assessment inventories have been proposed for use in the domestic violence context, but none have been properly validated and their predictive efficacy remains limited.  In this talk, I will discuss the methodological limitations of the existing literature, the challenges in producing defensible research in this domain, and the ethical implications of actuarial style risk assessment in the domestic violence context.  I hope to inspire discussion about ways to improve the quality of future research in this important domain.

Mar
13
Thu
EMVS: The EM Approach to Bayesian Variable Selection @ Sidney Smith Hall, Room 1074
Mar 13 @ 15:30 – 16:30

SEMINAR

Thursday March 13, 2014 at 3:30pm

Sidney Smith Hall, Room 1074   *Refreshments will be served at 3:15pm*

EMVS: The EM Approach to Bayesian Variable Selection

Veronika Rockova, University of Pennsylvania

 

Despite rapid developments in stochastic search algorithms, the practicality of Bayesian variable selection methods has continued to pose challenges. High-dimensional data are now routinely analyzed, typically with many more covariates than observations. To broaden the applicability of Bayesian variable selection for such high-dimensional linear regression contexts, we propose EMVS, a deterministic alternative to stochastic search based on an EM algorithm which exploits a conjugate mixture prior formulation to quickly and posterior modes.

Combining a spike-and-slab regularization diagram for the discovery of active predictor sets with subsequent rigorous evaluation of posterior model probabilities, EMVS rapidly identifies promising sparse high posterior probability submodels. External structural information such as likely covariate groupings or network topologies is easily incorporated into the EMVS framework. Deterministic annealing variants are seen to improve the effectiveness of our algorithms by mitigating the posterior multi-modality associated with variable selection priors.

The usefulness the EMVS approach is demonstrated on real high-dimensional data, where computational complexity renders stochastic search to be less practical. (Joint work with Edward George)

 

Seminars 2013-14 http://www.utstat.toronto.edu/wordpress/?page_id=18

 

 

Mar
17
Mon
Two-part Models for Semicontinuous Variables in Psychological Research @ Norm Endler Room (BSB 164)
Mar 17 @ 10:15 – 11:15

Quantitative Methods Forum @ Norm Endler Room (BSB 164)

Mar 17 @ 10:15 AM – 11:15 AM

Speaker: Dr. Dave Flora, York University
Department of Psychology

Title: Two-part Models for Semicontinuous Variables in Psychological Research

Abstract: “Semicontinuous” outcome variables arise regularly in psychological research, such as substance use research and developmental psychopathology, and applications in experimental psychology are also possible (e.g., response-time data). Such variables are characterized by a strictly non-negative continuous distribution coupled with a high frequency of observations equal to zero. Although models for zero-inflated count data are relatively well understood, models for continuous outcomes with a preponderance of zeros are less commonly used. I will describe both cross-sectional and longitudinal two-part models for such variables. Part 1 is a model for a binary outcome (whether a zero is observed) while Part 2 is a model for a continuous outcome, given that a non-zero value was observed in the first part. In the longitudinal case, model choice for Part 1 has implications for the interpretation of parameters in Part 2. These models will be illustrated with an application to adolescent alcohol use.

Mar
21
Fri
A Pipeline for High-Dimensional Time Course Gene Expression Data to Study Dynamic Network Responses to Viral Infections @ North Ross 638
Mar 21 @ 10:30 – 11:30
A Pipeline for High-Dimensional Time Course Gene Expression Data to Study Dynamic Network Responses to Viral Infections
Hulin Wu, Ph.D., Dean’s Professor
Department of Biostatistics and Computational Biology
Director, Center for Integrative Bioinformatics and Experimental Mathematics
University of Rochester School of Medicine and Dentistry
Email: Hulin_Wu@urmc.rochester.edu
A new pipeline for high-dimensional time course gene expression data is developed based on the concept of function data analysis (FDA) with a purpose to study dynamic network responses at gene level. The pipeline includes significant testing for dynamic response genes (DRGs), clustering gene response curves, constructing dynamic gene response networks using differential equation models, network feature analysis, dynamic system analysis, and biological annotations. Novel statistical methods and modeling approaches are developed for the pipeline, which include high-dimensional ODE model selection, parameter estimation, and dynamic system characteristic analysis. We illustrate the pipeline and the proposed methods using genome-wide time course gene expression data from mice and human subjects challenged by influenza viruses. Some interesting biological findings will be discussed.