I am currently an assistant professor in the School of Psychology at the Georgia Institute of Technology in Atlanta. Recently, I was an assistant professor of research at The University of Oklahoma Health Sciences Center (OUHSC) in the Center on Child Abuse and Neglect (CCAN). My dissertation was completed at The University of Oklahoma (OU). I was formerly working in the Human Dynamics Laboratory where Dr. Steven M. Boker was my primary advisor and where I conducted my Master's research. I sought bachelor's degrees in Mathematics and in Psychology from Michigan State University, and am now pursuing these interests further as faculty.
Selected Research Projects
It is worth noting that the below are generally old and out of date research projects. Please see my CV for details on more recent publications, conference presentations, and workshops.
Power Simulation of Block Randomized Trials
In association with David Bard at the University of Oklahoma Health Sciences Center, I have been conducting a large-scale simulation study aimed at analyzing the ability of randomized block designs to detect effects when they are present. The result of this simulation is file with 2.6x107 (26 million) rows. Analysis of data this large often presents its own challenges.
In collaboration with Joseph L. Rodgers, David E. Bard, William H. Beasley, and Kelly M. Meredith, I am part of the support for the R package NlsyLinks. This package uses information from the National Longitudinal Survey of Youth (NLSY) to create pairs of individual records with known heritability coefficients. The package also includes several utilities to model samples with known heritabilities by standard behavioral genetics methods. The project has a forum for users that I monitor associated with its R-Forge page.
Relations Between Hidden Markov Models and Structural Equation Models
My masters thesis work began to explore connections between hidden Markov models (HMMs) and structural equation models (SEMs). There are some surprising similarities to be found when these techniques are seen in a somewhat abstract setting. For instance, both maintain similar distinctions between latent/hidden variables and manifest/observed variables. Both can also be thought to have a measurement model that defines the mapping between latent and observed variables, and a structural model that defines how latent variables relate to one another. HMMs and SEMs are certainly not identical models, but thinking of them in an abstract way illuminates their similarities in ways that can lead to fruitful research avenues.
OpenMx: Free Statistical Software
I have been on the OpenMx Project core development team since the fall of 2008. The Mx statistical software, written by Michael Neale is a statistical package for doing Structural Equation Modeling. The OpenMx project is a complete rewrite of Mx making it Open Source, extensible, and generally ready for the next twenty years of statistics. I have contributed a number of features and manual pages. My main contributions have been threefold: (1) adding state space modeling (a time series method) in discrete and continuous time, (2) adding LISREL model specification to OpenMx to complement the existing RAM specification, and (3) adding weighted least squares estimation. Several of these features are described in the OpenMx 2.0 article in Psychometrika.
Patterns of Change in Panic Symptom Severity During Treatment
In collaboration with Shari Steinman and Bethany Teachman, I have been attempting to understand how levels of panic change over the course of 12 weeks of Cognitive Behavioral Group Therapy, and how individual differences in these patterns of change relate to symptom levels six months after treatment ends. Some of this work has been published and is available via this link.
Automatic Classification of Facial Expressions with Velocity Information
Working with Tim Brick and Jeff Cohn, we used the Cohn-Kanade database of Facial Action Units activated when making facial expressions to estimate the position, velocity, and acceleration of 68 points on the face. A portion of these estimates were then given to a support vector machine (SVM) so that it could learn the rules for making facial expressions. Finally, the SVM was tested to see how much additional information velocity contributed over position, and how much additional information acceleration provided beyond position and velocity. This project was written for and presented at the 2009 Affective Computing and Intelligent Interaction conference. The article should be available here or here.
Intraindividual Variability in Personality
As my undergraduate honors thesis at Michigan State University, I collected data on 10 individuals about 5 times per day for around 21 days using Palm Pilots and the Experience Sampling Program. We collected data on self-esteem, affect, and behavioral personality. I later modeled each individual and each trait separately as a damped and forced harmonic oscillator. My advisors were Ryan P. Bowles and M. Brent Donnellan.
The human face provides rich information about the affective state of an individual. Many researchers have investigated this link; however, relatively few have explored it dynamically by actively examining how the face moves instead of the static expressions it makes.
I have been a teaching assistant (TA) for undergraduate statistics, a lab instructor and TA for graduate statistics, and an online instructor for undergraduate statistics.