Jeffery Lab, Vanderbilt School of Nursing
Principal Investigator: Alvin Jeffery
Data: MIMIC III and MIMIC IV Database (Google Big Query)
- Developed a phenotyping pipeline by extracting a cohort of patients with positive viral pneumonia from MIMIC III and MIMIC IV database using ICD codes, lab values.
- Analyzed text data from discharge summaries of patients within the extracted cohort and used keywords to filter out negative cases.
- Applying unsupervised learning techniques (AGNES, K-Prototype, PAM) to identify latent groups of patients within the viral pneumonia cohort and understand the differences in outcomes (mortality rate, length of stay, intubation) within these groups.
Perception, Plasticity and Learning Lab , Vanderbilt University
Principal Investigator: Kari Hoffman
Data: Trial data collected via repeated experiments (19,000+ observation)
- Developed a feature engineering and modeling (logistic regression, Bayesian hierarchical modeling) pipeline.
- Analyzed predictor effects and interactions to understand how learning from errors or correct trials changes over the course of the learning curve.
- Modeled the trails that were performed under the influence of an allosteric modulator(drug) and noted the differences in the learning process while under the influence of said drug.
Data Science Institute + Gaydosh Lab, Vanderbilt University
(Submitted to Population Association of America conference)
Principal Investigator: Lauren Gaydosh
- Developed predictive modeling (distributed random forest, logistic regression with L1 regularization) pipelines to identify relevant predictors of self destructive behaviors (suicidal ideation, alcohol abuse, illegal drug usage and prescription drug abuse) using survey data from 9500 individuals over a timespan of 19 years (ADD Health).
- Conducted experiments to understand the differences in the relevance of the different predictors of despair within various domains of despair (behavioral, cognitive, individual etc).