I use statistics and data science to explore the changes in populations and population health. I am especially interested how law and policy impact these changes. Download a copy of my cv here, and explore my recent work below.
In 1999, the United States changed the way it assigned causes of death to individuals. This study analyzes the impact of that change. Using data from 23 million individual death certificates reported to the Centers for Disease Control, this study uses a series of Bayesian Hierarchical Models to sample posterior predictive data to identify the impact of the change. The poster above shows that the change erased a decline in the proportion of acute conditions (like pneumonia) relative to chronic conditions (like heart attacks). This was particularly true for blacks and women. This means some of the apparent “progress” against mortality from acute disease is really just a coding change. Accordingly, it is important to account for procedural changes when using cause of death statistics to measure population health. Download the poster or read a draft of the paper.
This study compares the relationship between paid malpractice claims by state, and deaths due to medical complications as reported to the Centers for Disease Control. I find more evidence of deterrence than malpractice. Download the graphic or read a draft of the paper. The infographic above implements newer statistical models based on event history analyses and uses rates.
Does experiencing a recession in young adulthood impact happiness later in life? Some social theories strongly suggest it should. Recessions in young adulthood making finding a job and getting started in life harder, and having a hard time in young adulthood might make someone more pessimistic about the future, or it might make someone more satisfied with less. In this study, I answer the question by combining pooled cross-sections and panel data from a number of general social surveys to answer the question with economic data from the Bureau of Labor Statistics. I find, as pictured above, that experiencing a short recession is associated with more happiness, but experiencing a longer recession is associated with less happiness, even decades later. Download the graphic. Read the working draft.
Dynamic Associations of Network Isolation and Smoking Behavior
Isolated adolescents have higher risks of depression, suicide, and substance abuse. However, researchers have found mixed results for isolation and smoking use. This study (1) separates isolation into three types, withdrawn isolates, avoided isolates, and the externally oriented; and (2) tracks associations of these isolation types and cigarette use across grades. I developed and implemented a four-dimensional, multivariate autoregressive latent trajectory model (a type of time-series structural equation model) and assisted my coauthors in writing the paper. We find that avoided isolation decreases odds of subsequent cigarette use, and cigarette use increases the odds of withdrawn isolation. Download the infographic or read a draft of the paper
U.S. Immigration Policy and Disability Among Mexican Migrants
This study uses Hierarchical Linear Models to estimate disability trajectories for Mexican Americans. We were particularly interested in finding out if the speed of disability growth for immigrants was different based on the strictness of the U.S. immigration law and policy when they arrived. The image above shows that more restrictive migration regimes are associated with larger numbers of disabilities. I analyzed the data, summarized the findings, and researched the legal background. My colleague Collin Meuller secured the data and wrote the discussion on work, gender and occupations. Download the graphic
This project rose out of my work at Duke’s Biodemography of Aging Research Unit (BARU). Using the Health and Retirement Study (HRS) waves 2-12 (1993-2015; 83,233 observations for 9,947 individuals), we test gene-by-environment associations with depressive symptoms. We measure the economic environment directly by labor force changes (retirement and unemployment), and indirectly by regional unemployment rates. We use individual fixed effect models to test whether changing economic climate enhances gene-environment associations (social trigger) or whether they diminish gene-environment associations (social push). We find evidence for both processes: gene-by-environment associations with retirement diminish over time, but gene-by-environment associations with regional unemployment rates rise over time. Read an extended abstract.
This project uses information from general social surveys to project changes in religious affiliation in the United States. We use both multivariate and univariate Bayesian models to generate posterior predictive data to project mortality, births, conversions, and apostasies for major religions in the United States. I analyzed the data and summarized the findings and my colleague Cyrus Schleifert placed it in the academic literature. The image above won a best poster award at the Population Association of American meeting. Download the poster.