I am an experienced statistician who is passionate about developing and implementing novel statistical inference procedures, sampling and experimental designs, and conducting rigorous statistical analyses to solve real world problems to inform public health.
My research work focuses on developing quantitative methods and study designs to address pressing questions in public health. My goal is to lend my expertise to design, develop and implement the most rigorous study designs and quantitative methods to answer research questions in the public health arena in a way that is unbiased, precise, interpretable, and actionable.
One of the most important areas of my research is the study of the homeless and of their risky behaviors and social networks. My contributions to this line of research span from the development of sophisticated sampling designs for obtaining large probability samples of homeless women, men, and youth from shelters, drop-ins, meal lines, and streets; to the design of a way to reduce respondent burden when asking homeless women to elicit their social networks; to more substantive contributions such as the study of the role of substance use in intimate partner violence among homeless women.
Another important area of my research is the development of rigorous analytical methods to estimate intervention effects in large randomized controlled mental health studies. I developed models and methods for demonstrating the effectiveness of cognitive behavioral therapy-based interventions implemented at the primary care level for panic and, more generally, anxiety patients.
From a more methodological point of view my research has focused on methods for causal inference in observational studies. Specifically, I work on optimizing the estimation of propensity score weights using machine learning algorithms such as boosting. I also investigate the use of boosting to estimate either the propensity score model or the outcome model, or both in a doubly robust context and compare its performance to other machine learning algorithms such as BART (Bayesian Additive Regression Trees). From an applied point of view, I have used these methods in a variety of settings: from estimating the effect of adding cognitive behavioral therapy to medications versus medications alone among anxious patients; to estimating the effect of parks and park improvements on physical activity; to estimating the effect of being a military spouse versus a civilian one on employment outcomes to name just a few.
Education
- Phd, University of Washington, 2000
- MS, University of Washington, 1997
- Laurea, Università Commerciale L. Bocconi, 1993
Social Justice
Dr Golinelli has worked on several research projects assessing the disparities in risky behaviors, health outcomes, and access to care of racial and sexual minorities.
Teaching
Dr Golinelli provides mentoring to both pre- and post-doctoral researchers. Before joining the University of Pennsylvania, Dr Golinelli taught both undergraduate and graduate statistics classes at the University of Southern California and the RAND Pardee Graduate School.
Research
Dr Golinelli conducts both methodological and applied research in the public health arena. She is particularly interested in studying the role of nurses in mitigating health outcome disparities and conducting evaluations of nursing programs in both experimental and quasi-experimental settings.
Opportunities to Learn and Collaborate at Penn Nursing
As a methodologist at CHOPR, Dr Golinelli collaborates with students, faculty and staff members on a variety of research projects evaluating the role of nurses on patient outcomes and reducing health disparities.
