Compiled by the Statistical Partnerships Among Academe, Industry, and Government Committee
Ruobin Gong, PhD Student, Harvard University
While an undergraduate at the University of Toronto, I was fascinated with cognitive behavioral research. The training I received was typical of an experimental psychologist, covering everything from running experimental subjects to writing fMRI processing scripts. Statistics as a discipline was first introduced to me that way. I didn’t make out the Cohen’s Ds, Newman-Keuls, or Latin squares at the time, but nevertheless grew a secret appreciation for the principled inferential reasoning statistics has to offer. In the fall of 2013, I took a leap of faith and began to pursue a PhD in statistics at Harvard University.
Life as a PhD student is undeniably rewarding, despite not being the easiest. Pressure, self-doubt, and sleep deprivation cannot offset the excitement when an algorithm finally works, a model fits, or a derivation checks out. But as a scientist-turned statistician, something still felt restless: the desire to be in sync with real-life data problems. It takes far more than statistical proficiency to carry out a project from end to end; an overseeing vision can only be gained through hands-on experience. So when I came across Data Science for Social Good (DSSG) at the beginning of 2015, I immediately knew I needed to take part in it. After a short application and a Skype chat with the director, Rayid Ghani, my summer was set and I was Chicago bound.
DSSG is The University of Chicago summer fellowship that works with government and nonprofit organizations to use data-driven methods to address their pressing concerns, employing data that are readily available but cannot be put to efficient use. Last summer, it created an unprecedented learning experience for an amazing cohort of 42 fellows, thanks to the dynamic nature between the projects and people. Representing diverse academic backgrounds, we the fellows worked on 12 projects for social good, tapping into every facet of the phrase’s meaning.
My project focused on identifying high-school students at risk of not graduating on time. Three U.S. public school districts partnered with us and provided historical student enrollment, academic performance, and discipline data, with which we built at-risk student early prediction models. Throughout the fellowship, my teammates and I operated on a high level of “vertical” autonomy, fulfilling tasks from directly communicating with partners to calibrate inferential goals and resolve data gaps to constructing our own query database to presenting and documenting our modeling work to both internal and external audiences.
Real-life data is incredibly messy—they look nothing like that design matrix X we see in a regression course. Moreover, whenever things can go wrong, they will go wrong. Throughout the past summer, I have been tremendously grateful for my teammates and peer fellows, for they gifted me with not only subject-matter knowledge for writing Python functions to fixing Git catastrophes, but also the art and practice of working in teams toward a mutual goal. I treasure the experience for all it has taught me and have no doubt it will prove crucial to forging successful collaborations with statisticians, scientists, and practitioners alike.
Tyler G. Kinzy and James P. Normington, Master’s Students, University of Minnesota
From James: Tyler and I are both pursuing master’s degrees in biostatistics at the University of Minnesota. Our department regularly notifies us about internship and employment opportunities, and the university has a job posting site that is regularly updated with both internal and external positions. We both started applying to a handful of applications, finally securing interviews and offers with the Allina Health Division of Applied Research (DAR).
All my previous work experience was in software, teaching, or running a register. I entered the graduate program with no research experience and felt lucky to find a position at DAR, a dynamic group of scientists, researchers, and statisticians who work alongside medical staff to improve the quality of care delivered locally and nationwide. It’s exciting to say the least.
I mostly work with emergency services data, which is concerned with pre-hospital care—care at the scene of incident, ambulance procedures, and dispatcher and paramedic performance. I primarily work in Stata, writing scripts to read in raw data, preparing it for analysis, and then conducting that analysis. Analyses I have conducted vary in complexity, but the usual suspects include Wilcoxon rank-sum tests, Chi-square tests, trend tests, and inter-rater reliability tests. Examples of current studies are evaluating cardiac arrest protocol adherence based on the American Heart Association’s 2015 guidelines, assessing paramedics’ well-being and job burnout via survey, and assessing the effectiveness of sepsis education. I also have taken the reigns in claiming lead analyst roles, writing methods sections, and preparing institutional review board packages.
Working at Allina has given me the research backbone necessary to enter either academia or industry, a backbone I severely lacked when beginning my graduate studies. As I reflect on my Allina experience, I am grateful for the opportunity to have an effect on meaningful research, the lessons I have learned, and the friendships I have made.
From Tyler: I differed from James in that I worked for five years in health care before enrolling as a master’s student. My previous work was in behavioral research, so the more clinical setting at Allina was a welcome change.
I was immediately impressed with DAR’s diversity in research areas and the autonomy afforded to investigators and interns alike. The research is unique in its almost immediate application to the health system—something I feel is sometimes lacking in more academic environments.
I mainly work on research that describes and aims to improve health care delivery in critical care units. Studies have included identifying risk factors for respiratory failure in elective orthopedic surgeries and predicting cerebral performance after receiving therapeutic hypothermia. Analyses vary from the purely descriptive to using generalized estimating equations models to account for time-clustered to building predictive models using penalized regression.
The internship has been a great experience, allowing me significant sway on study design, analysis, and dissemination of findings. And, of course, Allina has introduced me to many amazing individuals who I look forward to collaborating with in the future.
Steven Willke, Undergraduate Student, The Ohio State University
I am a senior at The Ohio State University studying economics and actuarial science. I came to Ohio State with aspirations to become an engineer. However, I realized after a year that only the quantitative challenge of engineering intrigued me, so I switched my academic focus to mathematics and economics. Very quickly, I knew I made the right decision.
I became even more confident after my experience working with Kelly Zou at Pfizer as a statistical intern the summer after my sophomore year. In the classroom, I learned important mathematical concepts, but I actually got to use things I learned with Kelly. I also was exposed to concepts I would not be taught until later in my collegiate career. For example, I was lucky enough to work on a project in which we researched the differences between fixed effects modeling and random effects modeling, which I didn’t learn in a more traditional sense until my junior year in an econometrics class. This project was eventually published in Applied Statistics in Biomedicine and Clinical Trials Design. Working on that project, along with the rest of my internship experience, reassured me that I wanted to do some sort of quantitative investigation work for a living.
I would encourage fellow undergraduate students interested in statistics or another quantitative field to seek out and take advantage of opportunities like the one I had. I remember feeling nervous about working with people with PhDs and master’s degrees while I was still an undergraduate student, but the people I worked with were very willing to teach me and give me work that interested and challenged me. Even though I commuted to work 90 minutes every day, I always looked forward to going into work. I really enjoyed the atmosphere of intelligent people working together using various statistical methods to assess the safety and effectiveness of a drug, to improve clinical and outcomes research study practices, and to assure methodologic quality in other processes at Pfizer.
I was extremely lucky to have accomplished individuals like Kelly to talk with about various fields and the options available to people with statistics expertise. I told her about my interest in an actuarial career and she gave me advice about how to be successful in the field. Although there is still a long road ahead of me to become a fully credentialed actuary, I am happy with the position I am in at this stage of my studies. I’m about to have my second internship for a major insurance company and my experience as a statistical intern has not only set me apart in obtaining those internships, but also has almost surely improved the likelihood of achieving my actuarial ambitions.