Research

For my master's thesis, I embarked on an ambitious research project to try and validate a new method of measuring the state-level prevalence of depression. I sampled 3.5 million texts on Twitter that express vocabularies associated with depression and anxiety (1st image). Then, I verified the presence of the sentiment using a deep-learning model. Grouping the final result by US states, and obtaining The Depression Index (purple line, 2nd image), I then cross-validate the index with the existing real index of suicide, drug- and alcohol-induced death using two-way fixed effects (TWFE) models. Results in the 2nd image show that spikes in depressive sentiments on Twitter are always followed by spikes in suicide, and drug and alcohol-induced death. The right side of the 3rd image shows which US states score higher on the measured depression index. TWFE results are also promising (not shown here).

The dashboard shows the result of my first project as a research intern at Fondazione Eni Enrico Mattei. The dashboard shows the trade network of a critical mineral for energy transition, and how risky are the trades, with adjustments to certain relevant components.