|
Home
| Contact | Publications
| Talks
SLAM Group | Research | CV | JHU
Official Page
Survival
Analysis (Biostatistics 140.641) — offered every year
This intermediate-level course introduces fundamental
concepts, theory and methods in survival analysis. The course emphasizes
statistical tools and model interpretations which are useful in medical
follow-up studies and in general time-to-event studies. The content
includes hazard functions, survival functions, types of censoring and
truncation, Kaplan-Meier estimates, log-rank tests and their
generalization. Parametric models and inference include likelihood
estimation and the exponential, Weibull, log-logistic and other relevant
distributions. Statistical methods and theory for the proportional hazard
model (Cox model) are discussed in detail with extensions to
time-dependent covariates. Clinical and epidemiological examples (through
class presentations) will be discussed and illustrated with various
statistical procedures in class and also through homework assignments.
Advanced Survival
Analysis (Biostatistics 140.741) — offered every
other year
This advanced course introduces statistical models and
methods useful for analyzing univariate and multivariate failure time data.
It extends the course of Survival Analysis (Biostatistics 140.641) to
topics on semi-transformation model, competing risks models, length-bias
and prevalent samplings, multivariate and frailty survival models, models
and methods for analyzing recurrent events data, and martingale theory for
counting processes. Emphases are placed on nonparametric and semiparametric
approaches for modeling, estimation and inferential results. Clinical and
epidemiological examples are presented in class to illustrate statistical
procedures.
Biomarkers, Risk Prediction and Precision
Medicine (Biostatistics
140.742)
— offered every
other year and jointly taught by Chattergee N.
and Wang MC
Statistical models are often used to predict or evaluate the
probability that an individual
with a given set of risk factors or biomarkers will
experience a clinical or disease outcome.
A risk prediction model can help in clinical decision making
and help patients make an
informed choice regarding their treatment. The predictive
ability of a model is usually
evaluated by the model’s ability to discriminate between low
and high risk patients, and an
assessment of calibration, that is, the agreement between
observed outcomes and predicted
outcome.
Part I of the course (Wang) will focus on characterization
and estimation of numerous
risks in relation to biomarker measurements and binary or
time-to-event outcome. Lectures
will be more toward method-theory discussion, with topics on
relative risk, absolute risk,
odds ratio and hazard ratio parameters in prospective and
case-control studies.
Part II of the course (Chattergee)
emphasized risk prediction with real data applications:
Discovery of risk factors —> Characterization of relative
risk —> Estimation of absolute risk
—> Evaluation of model calibration —> Evaluation of
public health utility (Chatterjee et al.
2016).
|
