I develop designs and methods of analyses to evaluate treatments in medicine, public health and policy (causal inference). The increased quality and number of available treatments, and increasing ethical and practical constraints, are transforming the field of intervention research: the factors of research interest are no longer (and correctly so) the same as factors that we can intervene on humans. To address this, we have been developing new designs and methods for partially controlled studies, that is, studies that explore the factors that can be controlled, in order to investigate the effects of the factors of research interest.

For example, even in the most reliable medical studies -- the ``randomized studies'', patients often do not comply with the assigned treatments and drop out. We have shown that the ``intention-to-treat method'', which has been widely used for those situations, is not suitable to generally estimate even the ``intention-to-treat effects'', and we have provided appropriate methodology. We have recently integrated this work with Don Rubin in a unifying statistical framework, ``principal stratification". Principal stratification allows researchers to formulate designs and address a challenging statistical problem with partial control: to find the degree to which the effect of a controlled treatment or factor on a main outcome is explained by the effect of the controlled treatment on the activation of intermediate causal pathways that are not directly controlled.

Principal stratification has now been applied in a broad range of areas, including HIV; cancer; ophthalmology; orthopedics; mental health; nephrology; surrogate endpoints; noncompliance with missing outcomes; and effects of vaccines on viral load for those infected.

CV

Biostatistics, Bayesian statistics; Causal inference; clinical trials; epidemiologic statistics; foundations of inference; longitudinal data analysis; missing data models

An, MW, Frangakis, CE, Musick, BS, and Yiannoutsos, CT. (2008). The need for double sampling designs in survival studies: an application to monitor PEPFAR. Biometrics (to appear).

Frangakis CE, Rubin DB, An, MW, and MacKenzie, E. (2007) Principal stratification designs to estimate input data missing due to death. Biometrics (with Discussion) 63, 641-662.

Li, F and Frangakis, CE (2006). Polydesigns in causal inference. Biometrics 62, 343–351.

Frangakis, CE, Brookmeyer, RS, Varadhan, R, Mahboobeh, S, Vlahov, D, and Strathdee, SA. (2004). Methodology for evaluating a partially controlled longitudinal treatment using principal stratification, with application to a Needle Exchange Program. Forthcoming in the Journal of the American Statistical Association .99, 239-249.

Barnard, J, Frangakis, CE, Hill, JL, and Rubin, DB. (2003). A Principal Stratification approach to broken randomized experiments: a case study of School Choice vouchers in New York City. Journal of the American Statistical Association (with discussion), 98, 299-323.

Frangakis, CE, and Rubin, DB. (2002). Principal stratification in causal inference. Biometrics, 58, 21-29.

Frangakis CE, Baker SG. (2001). Compliance subsampling designs for comparative research: estimation and optimal planning. Biometrics 57, 3, 899-908.

Frangakis CE, Rubin DB (2001). Addressing an idiosyncrasy in estimating survival curves using double sampling in the presence of self-selected right censoring (with discussion). Biometrics 57, 333-353.

Frangakis CE, Rubin DB (1999). Addressing complications of intention-to-treat analysis in the combined presence of all-or-none treatment-noncompliance and subsequent missing outcomes. Biometrika 86 365-379.