Web Applications

Economics Seminar Speaker Diversity


This slideshow requires JavaScript.

Jennifer Doleac (Texas A&M) and Elizabeth Pancotti along with a team of undergraduate assistants have collected years of historical data on invited speakers for economics department seminars at universities across the US. Invited seminars serve an important role in academic economists’ career trajectories, so they wanted to better understand the racial and gender composition of seminar speakers in aggregate, within specific departments, and within specific seminar series (e.g. labor economics or health economics). Myself and Kelsey Skvoretz built an R Shiny Dashboard for visualizing trends of women speakers and under-represented minorities. You can even compare universities and specific lecture series over time! More data is being added regularly, so keep checking in on this important project.

CHF Risk Prediction Tool (Demo)

Screen Shot 2019-03-20 at 4.35.07 PM

Working with a team of collaborators from the Regenstrief Institute, we constructed a random forest risk prediction model for congestive heart failure (CHF). We then built an application, leveraging RShiny and ggplot2, to visualize both population-level risk prediction (left) and the factors influencing individual patients’ risk predictions (right). This allows users to see not only an overall risk prediction but also the unique factors driving each patient’s predicted risk. The “individual patient” plot on the right dynamically updates when the user clicks on any point in the “population” plot on the left.

See this blog post for more information and background on the hackathon and app design.

Opioid Prescribing Rate Visualization Tool


In 2018, the Centers for Disease Control and Prevention released data on opioid prescribing rates at the county and state levels for 2006 through 2016. Working with Kevin Wiley, another doctoral student in health policy and management, we developed a Shiny app to visualize changes in prescribing rates over time. This app expands on the CDC data by allowing users to define regions and time frames for comparison. For more information, see this blog post.

EHR Survey Response Visualizer

Screen Shot 2019-03-26 at 2.03.11 PMStarting in 2016, a group of researchers from the Fairbanks School of Public Health at Indiana University have been studying the replacement of a legacy electronic health record system at large public health system in Indianapolis. As a part of this study, we have surveyed all EHR users across the health system for two years. After the data collection was completed. To assist in analyzing results, I built an RShiny app leveraging ggplot2 to visualize 17 survey items with Likert response scales. The app can display the results as percentages of the total, counts, or percentage of respondents in the top two (“agree” and “strongly agree”) categories. Results are plotted over six survey waves based on the waves selected by the user. The plot generation options pane can be seen here on the left.

I am not able to publish this application publicly due to the nature of the information contained in the survey results, however aggregated results from this study have been presented at several health informatics and health services research conferences.

%d bloggers like this: