Smith Center Fellowships and Other Education
The first of its kind in Boston, the Richard A. and Susan F. Smith Center for Outcomes Research in Cardiology places the Division of Cardiovascular Medicine
and BIDMC at the forefront of a field that will only become more critical in the years ahead as health care reform places a greater emphasis on providing high-quality, cost-effective care that is centered on the patient.
The Smith Center Fellowship
The Smith Center offers a fellowship training program for researchers interested in studying cardiovascular outcomes. The fellowship provides early career cardiologists with an opportunity to develop the expertise and skills needed to advance their portfolio in outcomes research.
The fellow will undertake formal training at Harvard T.H. Chan School of Public Health in research methodology, develop novel research ideas, apply for grants, conduct research and publish their findings. They will be expected to submit abstracts at national and international conferences, and develop relationships within the broader research community focusing on this specialized field of research in both academia and industry.
Statistical Analysis Software
As part of our research mission, we are making some statistical software available to the research community. The following text files are R programs that can be used to perform statistical analyses for the study of heterogeneity of treatment effect using data from randomized clinical trials. The statistical theories behind these programs are described in the following publications:
- Shen C, Li X, Jeong J. Estimation of treatment effect in a sub-population: an empirical Bayes approach. Journal of Biopharmaceutical Statistics. 2016;26:507-518.
- Taft L, Shen C. A non-parametric statistical test of null treatment effect in sub-populations. Journal of Biopharmaceutical Statistics. 2019;15:1-17.
Test_main.R is the main function to detect heterogeneity of treatment effect.
Test_functions.R includes all functions needed for Test_main.R.
Effect_estimation_main.R is the main function to estimate treatment effect in a subgroup.
Effect_estimation_functions.R includes all functions needed for Effect_estimation_main.R.
If you have feedback or questions regarding these programs, you can reach our Statistical Director, Dr. Changyu Shen, at firstname.lastname@example.org.
Statistical Education Series
Each academic year, faculty at the Smith Center offer an 8-week long statistical education series for cardiovascular medicine fellows interested in pursuing research. This series builds a strong conceptual foundation in statistical inference, familiarizes participants with the principles of basic statistical analysis tools, and applies these principles within three major areas: randomized trials, observational studies, and prediction.
Course Summary, fall 2017
- Basic statistical concepts I (population, sampling, parameters,
statistics, hypothesis testing, estimation, sampling distribution)
- Basic statistical tools (univariate, bivariate and multivariate;
parametric versus non-parametric)
- Randomized trials (rationale, design, sample size, analysis)
- Randomized trials (benefit-risk trade-off)
- Observational studies (confounding, potential outcomes, design)
- Causal inference (propensity score, outcome regression, matching,
- Predictions (prediction model, model fitting, bias-variance trade-off)
- Predictions (metrics of performance, estimation of performance)
The objective of this education series is to build a strong conceptual foundation in the philosophy of statistical inference and its application to different analytical situations in the real world during the design and analysis stages. The educational focus is on concepts (“what” and “why”) instead of implementation (“how”).
Upon completion of this series, the audience is expected to have clear understanding of the basic principles of different statistical techniques, when they should be applied, and how to interpret the results of these analyses. The audience is expected to be able to understand medical research articles with heavy statistical components and communicate proactively with professional statisticians.