Fall 2019
The Quantitative Psychology colloquium series meets weekly in the Autumn and Spring semesters. The primary activity is presentation of ongoing research by students and faculty of the quantitative psychology program. Often, guest speakers from allied disciplines (e.g., Education, Statistics, and Linguistics) within and external to The Ohio State University present on contemporary quantitative methodologies. Additionally, discussions of quantitative issues in research and recently published articles are often conducted during colloquium meetings.
Faculty coordinator: Dr. Jolynn Pek
Venue: 35 Psychology Building
Time: 12:30-1:30pm
August 26, 2019
Organizational Meeting
*pizza and drinks will be served!
September 2, 2019
Labor Day
September 9, 2019
Speaker: Joonsuk Park
Department of Psychology, The Ohio State
Title: Using Sequential Sampling Models as Measurement Tools: Pitfalls and Recommendations
Abstract: Sequential sampling models are often used as measurement tools in psychology. However, the practices regarding the use of them have received relatively limited attention. Specifically, identifiabilities of the models and the statistical power of follow-up testing procedures have not been considered in detail. In this presentation, I do the following two things to fill the gap in the literature. First, I conduct identifiability analyses of the two popular sequential sampling models, namely the Diffusion Decision Model (DDM) and the Linear Ballistic Accumulator (LBA). I show that some version of the diffusion decision model (DDM) are unidentifiable, even when multiple experimental conditions are employed. Second, I demonstrate that the use of t-tests while comparing parameter estimates lowers the statistical power, and employing alternative procedures that explicitly model uncertainties about the parameter estimates, such as meta-regression or hierarchical Bayes, is better in terms of statistical power. Finally, recommendations for substantive researchers are provided.
Discussant: Seo Wook Choi
September 16, 2019
Speaker: Dr. James Austin
College of Education and Human Ecology, The Ohio State
Title: Professional Standards in Testing and Assessment
Abstract: In this presentation I attempt to integrate the recent revision of the Standards for Educational and Psychological Testing (AERA, APA, & NCME, 2014) with a resurgence of interest in validity (Markus & Borsboom, 2013; Newton & Shaw, 2014). For the first time, psychometricians outside academia participated on the revision committee. I begin with a brief overview of professional organizations including the Association of Test Publishers (ATP) and its divisional structure, the International Test Commission (ITC) and its handbook publication project, and the National Council on Measurement in Education (NCME). Then, focus on testing standards and guidelines emphasizes the dominant 2014 Standards as well as additional sets of standards promulgated by the International Test Commission. Argument-based validation is discussed in the context of the Mislevy (ECD) and Kane models. I conclude with some purposes and practices of professional standards – because they are voluntary, issues of alignment among standards and compliance-commitment by practitioners are important. Alignment may be achieved by revising standards post-2014 (AERA-APA-NCME). Two examples include the Institute of Credentialing Excellence (ICE) releasing its revision of Standards for the Accreditation of Certification Programs (2014) and the Society of Industrial-Organizational Psychology releasing the Principles for Validation and Use of Personnel Selection Procedures (2018).
James T. Austin earned his Ph.D. in Industrial/Organizational Psychology from Virginia Tech in 1987, then completed a Postdoctoral Traineeship in Quantitative Psychology at the University of Illinois in Urbana-Champaign (1987-89). He served as a visiting faculty member at the University of Illinois during his postdoctoral traineeship, and at New York University (1990-91), and was on the regular faculty at The Ohio State University from 1991-1997. In 1997, he moved to the Center on Education and Training for Employment in the College of Education and Human Ecology at The Ohio State University. He is currently a Senior Research Specialist serving as Program Lead of CETE Assessment Services. He works on projects focused on measurement of occupational-technical knowledge and skill; most involve developing and maintaining testing-assessment systems for educational and business client organizations. His research interests include statistical and methodological topics as well as testing and job performance measurement – increasingly related to competency models.
September 23, 2019
Speaker: Dr. Dylan Molenaar
Department of Psychology, University of Amsterdam
Title: A Semi-Parametric Moderated Factor Analysis Approach to Test for Measurement Invariance across a Continuous Variable
Abstract: Effort has been devoted to the development of moderated factor models in which the traditional factor model parameters are allowed to differ across a moderator variable. These models are valuable as they enable tests on measurement invariance across a continuous background variable. However, moderated factor models require the specification of a parametric functional form between the factor model parameters and the moderator variable while, in some situations, it is unclear what functional form to assume. Therefore, in the present paper, a semi-parametric moderated factor modeling approach is presented in which no assumption concerning the functional form between the moderator and the model parameters are imposed. In a simulation study, the semi-parametric moderated factor model is shown to be viable in terms of parameter recovery, true positive rates, and false positive rates. In addition, the model is applied on a real data set pertaining to intelligence.
Dr. Dylan Molenaar is an assistant professor at the department of psychology. Dylan received his PhD degree in psychology from the University of Amsterdam in 2012 (cum laude). Title of his thesis was “Testing distributional assumptions in psychometric measurement models with substantive applications in psychology”. His research interests are item response theory, factor analysis, response time modeling, intelligence, and statistical modeling of genotype by environment interactions. More information on his work is available in the links below:
https://www.uva.nl/en/profile/m/o/d.molenaar/d.molenaar.html
http://www.dylanmolenaar.nl/
Wherry Lecture
October 3, 2019: 3:00 – 4:00 pm
209 W 18th Ave. Room EA 170
Speaker: Sharon Bertsch McGrayne
Title: Bayes’ Rule: An underground theory that took over the world
Abstract: Bayes’ rule is ubiquitous today in many fields of endeavor, from underwater sea searchers to predicting the outcome of elections. but for most of the 20th century, it was used only in desperation because of the withering opposition of mathematical statisticians. Dr. McGrayne’s talk will take us through some oft he highlights of this early period, including Turing’s decoding of the German Enigma messages during World War II and what changed to bring about its popularity today.
Sharon Bertsch McGrayne is the author of highly-praised books about scientific discoveries and the scientists who make them. She is interested in exploring the cutting-edge connection between social issues and scientific progress – and in making the science clear and interesting to non-specialists. Thus, her first book dealt with changing patterns of discrimination faced by leading women scientists during the 20th century. Her second book portrayed a group of chemists and the interplay between science, the chemical industry, the public’s love of creature comforts, and the environment. Her latest book tells how an 18th century approach to assessing evidence was ignored for much of the 20th century before – in an overnight sea change – it permeated our modern lives.
October 7, 2019
Speakers: Dr. Andrew Hayes & Jacob Coutts
Department of Psychology, The Ohio State University
Title: Omega, Not Alpha for Estimating Reliability. But…
Abstract: Cronbach’s alpha is the dominant measure of reliability used when trying to measure or convey the amount of random measurement error that exists in a sum score or average resulting from the use of a multi-item measurement scale. Yet methodologists have warned that alpha is not ideal relative to its more general form often called McDonald’s omega. Among other reasons, that the computation of omega is not available in such programs as SPSS and SAS and requires items loadings (and perhaps item error variances) from a confirmatory factor analysis has probably hindered its widespread adoption. After a bit of discussion of alpha versus omega, I briefly illustrate the computation of omega using two popular SEM programs (Mplus and Amos) and the MBESS package for R. The bulk of the remaining part of my presentation focuses on a new and easy-to use macro for SPSS (a version will be available for SAS) that calculates omega in two ways, neither of which relies on the estimation of loadings or error variances using confirmatory factor analysis. I show by way of 17 examples that it seems to produce estimates of omega that are similar to if not largely identical to what is generated using CFA-based estimates of item loadings and error variances. I will then ask whether it actually even matters in day-to-day practice which measure we use.
This presentation (and the corresponding paper) was motivated by a paper coming out in “Measurement” by Greg Hancock and Ji An. If you want to read their paper, they (or someone) has made it available at https://www.researchgate.net/publication/335144823. I will tell you why their algorithm, while reasonably effective and also implemented in our macro, is not necessary to get omega out of SPSS. Indeed, we can do it better than them and still without requiring a program that can conduct CFA.
October 14, 2019
Speaker: Dr. Blake McShane
Kellogg School of Management, Northwestern University
Title: Circulated internally
Abstract: Circulated internally – please contact the coordinator for information.
Dr. Blake McShane joined the marketing faculty at the Kellogg School of Management in 2010 as a Donald P. Jacobs Scholar. He has developed and applied statistical methodology to topics ranging from optimizing internet ad-serving algorithms to forecasting home runs in baseball. His specific research interests include Bayesian hierarchical modeling, statistical learning, and generalized Markov models.
October 21, 2019
Speaker: Yiyang Chen
Department of Psychology, The Ohio State University
Title: Progressive ratio task: Bayesian hierarchical model and group differences
Abstract: The progressive ratio task is an experimental schedule aimed to measure “motivation”, i.e. the maximum amount of effort that a subject is willing to exert for a fixed amount of reward. The task has been used to quantify human motivational deficits, such as amotivation in schizophrenia patients (Wolf et al., 2014). I build a Bayesian hierarchical model to investigate the effects of multiple covariates on motivation, including the reward level, effort level and time elapsed in the experiment. By applying the model to Wolf et al. (2014)’s data, I find the elapsed time to be a possible factor related to schizophrenic amotivation, where schizophrenia patients appear to lose motivation faster as time elapses than neurotypical controls.
Discussant: Joonsuk Park
October 28, 2019
Speaker: Sarah Worch
College of Education and Human Ecology, The Ohio State University
Title: Family History of Diabetes or Heart Disease and Current Levels of Differentiation of Self in Couple Members: An Exploratory Study
Abstract: This presentation will report on a study that tested Bowen family systems theory’s proposition that emotional and physical symptoms are a reflection of the family’s multigenerational emotional process, and that differentiation of self is relationally linked to this process. To test this proposition, the current study examined the effect of having a family history of diabetes or heart disease on current couple members’ levels of interpersonal differentiation of self. The results of the Actor-Partner Interdependence models used in the current study will be discussed along with implications of the findings.
Discussant: Ivory Li
November 4, 2019
Departmental Colloquium
Speaker: Dr. Lisa Feldman Barrett
Department of Psychology, Northeastern University
Department of Psychiatry, Massachusetts General Hospital, Boston, Massachusetts;
Athinoula A. Martinos Center for Biomedical Imaging, Massachusetts General Hospital,
Boston, Massachusetts
Title: From essences to predictions: A brain-based understanding of emotion.
Abstract: This talk will explore a collection of scientific findings about emotions whose implications seem to defy 2000 years of common sense. We’ll then explore a new set of hypotheses about the nature of emotion which suggests that understanding ad hoc concept construction and bodily regulation, within a continually predicting brain, are key to understanding how actions are regulated and experiences are assembled. Theory and research will be organized around two themes: (1) the shift from taxonomic thinking (the belief that an emotion category has a particular facial expression and autonomic fingerprint) to population thinking (evidence that emotion categories consist of unique instances that are tailored to the specifics of the immediate situation); (2) the shift from essentialism (the belief that all instances of an emotion category share an underlying neural circuit) to degeneracy (evidence that instances of an emotion category are constructed as different configurations within the brain’s predictive architecture). Taken together, these themes represent a novel framework for the science of emotion in which brain, body, and context are unified.
Dr. Lisa Feldman Barrett is a University Distinguished Professor of Psychology at Northeastern University, with appointments at Harvard Medical School and Massachusetts General Hospital. In addition to the book How Emotions are Made: The Secret Life of the Brain, Dr. Barrett has published over 200 peer-reviewed, scientific papers appearing in Science, Nature Neuroscience, and other top journals in psychology and cognitive neuroscience, as well as six academic volumes published by Guilford Press. She has also given a popular TED talk.
Dr. Barrett received a National Institutes of Health Director’s Pioneer Award for her revolutionary research on emotion in the brain. These highly competitive, multimillion dollar awards are given to scientists of exceptional creativity who are expected to transform biomedical and behavioral research. She also received a Guggenheim Fellowship in 2019.
Among her many accomplishments, Dr. Barrett has testified before Congress, presented her research to the FBI, consulted to the National Cancer Institute, appeared on Through The Wormhole with Morgan Freeman and The Today Show with Maria Shriver, and been a featured guest on public television and worldwide radio programs. She is also an elected fellow of the American Academy of Arts & Sciences and the Royal Society of Canada.
November 11, 2019
Veteran’s Day
November 18, 2019
Speaker: Dr. Yunzhang Zhu
Department of Statistics, The Ohio State University
Title: A new inference method for high dimensional models
Abstract: In this talk, I will present a new methodology of statistical inference for high dimensional statistical models. For high dimensional models, the common practice of inference uses either a regularized model, as in inference after model selection, or bias-reduction known as “debias”. While the first ignores statistical uncertainty inherent in regularization, the second reduces the bias inbred in regularization at the expense of increased variance. I will present a constrained maximum likelihood method approach for hypothesis
testing involving unspecific nuisance parameters, with a focus of alleviating the impact of regularization on inference. Particularly, for general composite hypotheses, we unregularize hypothesized parameters whereas regularizing nuisance parameters through an $\ell_0$-constraint controlling the degree of sparseness. This approach is analogous to semi-parametric likelihood inference in a high-dimensional situation. On this ground, for the Gaussian graphical model and linear regression, we establish the asymptotic distribution of the constrained likelihood ratio test statistic, permitting parameter dimension increasing with the sample size. Interestingly, the corresponding limiting distribution is chi-square or normal, depending on if the co-dimension of a test is finite or increases with the sample size, which goes beyond the classical likelihood ratio test. Finally, I will present some numerical examples to demonstrate that the proposed method performs well against its competitors in various scenarios.
November 25, 2019
Speaker: Seo Wook Choi
Department of Psychology, The Ohio State University
Title: Bayesian joint modeling approach of brain network and response time
Abstract: There have been researches to find out how the brain network is structured. Using graph theory, there is some success of classifying the normal brain to the ill brain.However, the relationship between the brain network and behavior, such as response time and accuracy has not studied much. The current approach of model-based cognitive neuroscience has much effort to find out a connection between neural and behavioral data formally. It is still limited to regional brain activation to behavioral models. I am suggesting the way to connect the network model of the brain to the behavior model. Under the Bayesian joint modeling framework, the behavioral sub-model (Linear ballistic accumulator) and the neural (fMRI) sub-model (Economical clustering model) are combined. The simulation result and real data application will be discussed.
Discussant: Sarah Worch
December 2, 2019
Brief presentations on external talk
*pizza and soda will be served
Jacob Coutts
Sarah Worch
Seo Wook Choi
Yiyang Chen