FRG Informal Talk Series: Andrew Liu and Wan Zhang

2:00-2:15pm Lighting talk by Andrew Liu

Title: Public Perceptions of Synthetic Cooling Agents on Twitter

Abstract: Amid a potential menthol ban, electronic cigarette (e-cigarette) companies are incorporating synthetic cooling agents like WS-3 and WS-23 to replicate menthol/mint sensations. This study examines public views on synthetic cooling agents in e-cigarettes via Twitter data. From May 2021 to March 2023, we used Twitter Streaming API, to collect tweets related to synthetic cooling agents with keywords such as ‘WS-23,’ ’ice,’ and ’frozen.’ Two deep learning roBERTa models, classified attitudes expressed in tweets about synthetic cooling agents and identified e-cigarette users. The BERTopic deep-learning model identified major topics of positive and negative tweets. Two proportion Z-tests were used to compare the proportions of positive and negative attitudes between e-cigarette users (vapers) and non-e-cigarette-users (non-vapers). Of 6,940,065 e-cigarettes related tweets, 5,788 non-commercial tweets related to synthetic cooling agents. The longitudinal trend analysis showed a clear upwards trend in discussions. Vapers posted most of the tweets (73.05%, 4,228/5,788). Nearly half (47.87%, 2,771/5,788) held a positive attitude toward synthetic cooling agents, which is significantly higher than those with a negative attitude (19.92%,1,153/5,788) with a P-value < 0.0001. The likelihood of Vapers expressing positive attitudes (60.17%, 2,544/4,228) was significantly higher (P < 0.0001) than that from non-vapers (14.55%, 227/1,560). Conversely, negative attitudes from non-vapers (30%, 468/1,560) was significantly (P < 0.0001) higher than vapers (16.2%, 685/4,228). Prevalent topics from positive tweets included “enjoyment of specific vape flavors,” “preference for lush ice vapes,” and “liking of minty/icy feelings.” Major topics from negative tweets included “disliking certain vape flavors” and “dislike of others vaping around them.” On Twitter, vapers are more likely to have a positive attitude toward synthetic cooling agents than non-vapers. Our study provides important insights into how the public perceives synthetic cooling agents in e-cigarettes, insights that are crucial for shaping future FDA regulations aimed at safeguarding public health.

2:15-2:50pm Informal talk by Wan Zhang

Title: Regularization of BELIEF under Smoothness Interpretations

Abstract: The binary expansion linear effect (BELIEF) is an alternative framework to the popular GLM framework by taking advantage of a parallelism between bi-variate binary distributions and bi-variate normal distributions. One major challenge in the BELIEF framework is the exponential growth in dimensionality. To ensure statistical estimation and computational scalability, we propose a novel definition of smoothness for binary interactions. We have proved that this correspondence between the smoothness of each marginal binary interaction and the interaction itself is a one-to-one function. Additionally, we have shown that in higher dimensions, the smoothness of a binary interaction can be expressed as a product of marginal binary interactions. Based on these observations, we propose to model the dependency that is smooth with a generalized LASSO model with larger penalty on less smooth terms. The numerical studies show that the smooth LASSO for BELIEF provides several key advantages in interpretability and effectiveness.