Using Cluster Analysis to Understand the Relationship Between Emotion Regulation & Depression Symptoms (DSI-SRP)
This DSI-SRP fellowship funded Emma Boldwyn to work in the laboratory of Autumn Kujawa in the Department of Psychology and Human Development during the summer of 2022. Emma is a senior with majors in Cognitive Studies and Child Development.
Emma used data from a project that examines adolescents’ depression symptoms in conjunction with neurological factors. Specifically, data was pulled from the emotion regulation task, or ERT, which is a computer task that participants complete in the lab that collects continuous EEG data. Participants are taught strategies like cognitive reappraisal and asked to either reappraise or just look at an image on screen. One component of interest that we can look at using the ERT is the late positive potential, or LPP. This is an event related potential that occurs at least 300 milliseconds (between 450-1000 milliseconds) after a stimulus is presented. Previous research has found differences in the LPP among adults and adolescents with and without depression and other internalizing disorders. It’s also been associated with an increased risk for developing psychopathology later on. Emma used principal components analysis (PCA) and cluster analysis to work backwards and first identify unnamed factors that were significant within the EEG data, then group participants based on those factors and examine traits to see what they had in common. PCA makes data easier to analyze and interpret while also minimizing information loss and enhancing accuracy. Cluster analysis is a useful tool for looking at underlying factors in a dataset and examining different group traits based on performance on a task. PCA results indicated that there were likely 2 main significant components to focus on helped validate the decision to fit the data into 2 cluster groups, and that a main differentiation between those 2 cluster groups would be data from participants who were sensitive to emotional content in general versus participants who were particularly sensitive to reappraisal. Cluster analysis confirmed that 2 clusters was the best fit to the data. She ran additional descriptive analyses in SPSS, and only a significant difference between clusters in the look and reappraise conditions, where cluster 1 was more responsive. There was no association between clusters and self-report depression symptoms. Next steps could be examining clustering groups in relation to parental depression risk and history.
In addition to receiving support through a DSI-SRP fellowship, this project was supported and facilitated by the DSI Data Science Team through their regular summer workshops and demo sessions.