I expected to face roadblocks and challenges when I signed up to be involved with research, but it doesn’t make them any less sensitive when you are faced with them. The primary project I intended to work on this semester has been stymied for the past few months.
I have been analyzing data related to decision-making and mental health in childhood from the National Institutes of Mental Health’s Adolescent Brain and Cognitive Development (ABCD) study. I’m interested in a measure of decision-making known as delayed discounting which is essentially a measure of immediate gratification. Every few years, children in the study are given a task involving a series of varying binary choices between smaller, more immediate rewards and those of greater amounts later in the future. Subsequent options are altered based on an individual’s choices. For example, if a child chooses the sooner option, their next set of choices will include the same sooner amount with an increased later amount.
Typically, the repeated sequence of decisions is represented by a simple function and plotted on a curve such as the one above. However, these data are often noisy, and classical frequentist statistics produce inconsistent, counterintuitive results. Therefore, my project is to apply tools from Bayesian statistics to compute delayed discounting scores. Unfortunately, because there are data from over 8,000 participants, applying Bayesian inference has been a time-consuming process. The sheer size of the models requires weeks to complete, and we have had to re-run the models due to computational error multiple times now.
While working through these analyses, I relied heavily on support from my mentor and our lab’s data technician, who helped troubleshoot my issues with computing delayed discounting scores at scale. I now understand why big data projects can be so time-consuming and research-intensive. As frustrating as computing complications can be, the challenge has certainly fostered growth as well. The knowledge I’ve obtained through the project has opened the door to working on other projects using large datasets. Due to the roadblocks I’ve faced in moving this project forward, I have spent the semester working on a few other projects related to poverty and decision-making. I’m especially excited about an online experiment I’m designing which I plan to spin into a Bachelor of Philosophy thesis by the end of the academic year.