From Regression Analysis to Deep Learning: Development of Improved Proxy Measures of State-Level Household Gun Ownership

• Published in: Patterns (New York, N.Y.) Vol. 1; no. 9; p. 100154
• Main Authors: Gomez, David Benjamin, Xu, Zhaoyi, Saleh, Joseph Homer
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 11.12.2020; Elsevier
• Subjects: deep learning; firearm prevalence; gun ownership; proxy measures; regression analysis; deep learning; proxy measures; gun ownership; DSML 5: Mainstream: Data science output is well understood and (nearly) universally adopted; firearm prevalence; regression analysis
• ISSN: 2666-3899; 2666-3899
• DOI: 10.1016/j.patter.2020.100154

In the absence of direct measurements of state-level household gun ownership (GO), the quality and accuracy of proxy measures for this variable are essential for firearm-related research and policy development. In this work, we develop two highly accurate proxy measures of GO using traditional regression analysis and deep learning, the former accounting for non-linearities in the covariates (portion of suicides committed with a firearm [FS/S] and hunting license rates) and their statistical interactions. We subject the proxies to extensive model diagnostics and validation. Both our regression-based and deep-learning proxy measures provide highly accurate models of GO with training R2 of 96% and 98%, respectively, along with other desirable qualities—stark improvements over the prevalent FS/S proxy (R2 = 0.68). Model diagnostics reveal this widely used FS/S proxy is highly biased and inadequate; we recommend that it no longer be used to represent state-level household gun ownership in firearm-related studies. [Display omitted] •We develop deep-learning and regression-based proxies of state-level gun ownership•Both new proxies significantly outperform existing ones•We found that the widely used FS/S proxy is highly biased and inadequate•We recommend FS/S no longer be used to represent state-level gun ownership Data on state-level household gun ownership is largely missing in the United States, yet this variable is essential for firearm-related research and policy development. In the absence of gun ownership data, researchers and policy-makers have had to rely on proxy measures to represent this indispensable variable. Historically, the portion of suicides committed with a firearm has been regarded as the best proxy measure of gun ownership. In this work, we challenge this notion and develop two significantly improved proxy measures using first, traditional regression analysis and then the tools of deep learning. Our new proxy measures are both highly accurate and easy to obtain, and they can be used for a variety of purposes in cross-sectional studies of firearm-related violence at the state level. In the absence of direct measurements of state-level household gun ownership, the quality and accuracy of proxy measures for this variable are essential for firearm-related research and policy development. In this work, we develop two significantly improved proxy measures of state-level household gun ownership via two methods: first using regression analysis, then using deep learning. We subject the new regression-based and deep-learning proxies to critical examination, and we benchmark our new proxies against existing ones for accuracy and validity.