EPI289. Causal inference

Instructor: Miguel Hernán
Harvard School of Public Health, Spring semester

Causal inference from observational data is a key task of epidemiology and of allied sciences such as sociology, education, behavioral sciences, demography, economics, health services research, etc. These disciplines share a methodological framework for causal inference that has been developed over the last decades. EPI289 presents this unifying causal theory and shows how epidemiologic concepts and methods introduced in EPI201 and EPI202 can be understood within this general framework. The course emphasizes conceptualization but also introduces statistical models and methods for time-varying exposures, which will be extensively studied in EPI204 and EPI207.

Course Link: http://isites.harvard.edu/course/hsph-epi289-01

EPI207. Advanced Epidemiologic Methods

Instructors: James Robins and Miguel Hernán
Harvard School of Public Health, Fall semester

This course provides an in depth investigation of statistical methods for drawing causal inferences from observational studies. Informal epidemiologic concepts such as confounding, comparability, overall effects, direct effects, intermediate variables, and selection bias are formally defined within the context of a counterfactual causal model. Methods for the analysis of the causal effects of time-varying exposures in the presence of time dependent covariates that are simultaneously confounders and intermediate variables are emphasized. These methods include g-estimation of structural nested models, inverse probability weighted estimators of marginal structural models, and g-computation algorithm estimators. As a practicum, students reanalyze data sets using the above methods.

Course Link: http://icommons.harvard.edu/~hsph-epi207-01/

BIO288. Semiparametric Methods for Analysis of Missing and Censored Data

Instructor: Andrea Rotnitzky
Harvard School of Public Health, Spring semester

The goal of the course is to provide a comprehensive discussion of optimal estimation techniques for low dimensional parameters of semiparametric models (i.e. models with infinite dimensional nuisance parameters) for complex longitudinal data subject to informative censoring or missingness. The course will start with the discussion of the fundamental notions and results of semiparametric theory: pathwise derivatives, tangent space, semiparametric variance and information bounds, and influence functions. It will then provide a general estimating function methodology for locally semiparametric efficient estimation and doubly robust estimation under data that are coarsened at random. This general methodology will then be applied to derive locally efficient doubly robust estimators of 1) regression parameters in multivariate generalized linear models subject to missing at random data, 2) the survival function of an endpoint subject to dependent right censoring, 3) the quality of life adjusted survival time subject to dependent right censoring 4) the survival function of multivariate failure time data subject to univariate (dependent) censoring, 5) Cox regression parameters based on dependent right censored data and 6) smooth parameters of the distribution of a time to an endpoint outcome based on current status data and interval censored data.

Course Link: http://my.hsph.harvard.edu/icb/icb.do?course=hsph-bio288-01&pageid=tk.page.hsph-bio288-01.generalinfo

Causal inference

Instructor: Miguel Hernán
Erasmus Summer programme, Rotterdam, The Netherlands – AugustKarolinska Institutet, Stockholm, Sweden – December (not offered every year)