Please consult my CV for my most recent projects.
We often think about processes in a continuous-time way. If something makes me upset, it logically follows that the greatest intensity of those negative emotions will be immediately following the triggering stimulus. However, as time passes on, it is likely that I will "get over it" and the impact of the singular moment will have a progressively weaker effect on my current mood. Inherently, this is a continuous process; however, the means by which we typically model individual dynamics is not reflective of this thinking. That is, we primarily fit models in discrete-time.
One of the major hurdles for fitting continuous-time models has been the difficulty in setting up the models and getting them to converge. This--coupled with the rise of high-dimensional dynamic network models--has made it very difficult to fit continuous-time models to large sets of data due the need to parameterize a large dimensional dynamics matrix. Newer techniques are emerging such as the integration of regularization to continuous-time structural equation models (CT-SEM) that are promising for variable selection by means of penalization.
In my dissertation work, I propose an algorithmic procedure for automating the fitting of continuous-time dynamic networks by means of model modification. Stay tuned for the results :)
My current work focuses on the development of methods for identifying nomothetic trends in idiographic processes. In the more common tongue, I study how we can utilize information from intraindividual analyses (i.e., N = 1 models) to gain insight into prototypical dynamics in groups or "clusters" of individuals. This serves as an alternative to purely nomothetic analyses which assume that individuals in and of themselves are derived from a single population, possess identical coefficient models, etc.
A talk I presented regarding some of my work can be found here and the preprint for our paper in the Special Issue of the European Journal of Psychological Assessment (EJPA) and European Journal of Personality (EJP) can be found here.
In psychological research, IRT is often seen as a cumbersome and difficult process to undertake. To address these concerns, a portion of my research is concerned with explicitly comparing outcomes produced by classically derived assessments of psychological constructs (i.e., factor scores and/or ordinary summed scores) to those produced by scores derived from IRT. Ultimately, the goal of this facet of my research is to elucidate upon the tangible benefits of IRT over classical methods to justify the greater 'work load' and 'difficulty' associated with IRT methods.
My thesis and subsequent published work pertained to the utilization of IRT test equating procedures--specifically the Haebara characteristic curve method--to render the scores of multiple informants comparable across a longitudinal timeframes of 8-years (i.e., 9-measurement periods).
This research has manifested itself as an iterative selection algorithm that takes item responses and generates optimal tests based on model fit. This algorithm that we've developed reduces the time to construct scales while preserving the 'locality of information'. This algorithm currently works on 2PL and 3PL IRT frameworks. Plans are to extend this algorithm to work with the polytomous models of IRT as well.
Past Work
A substantial amount of literature currently exists suggesting the relative ineffectiveness of negatively worded items (NWIs) in psychological assessment. My research has focused on utilizing IRT test equating procedures to render positively worded and negatively worded items equivalent with regard to construct validity.
I am blessed to be able to collaborate with the following extraordinary people/researchers:
Deshawn Sambrano
Netasha Pizano
My MA advisor, Dr. Kathleen Preston
My thesis committee member, Dr. Jessica Tessler
The Fullerton Longitudinal Study