Core Seminar Series

Core Seminars

The Applied Biostatistics Seminar Series consists of a series of talks with the primary purpose of furthering statistical knowledge on an applied level. Talks will focus on advances in biostatistical methods and statistical programming techniques and their translation into addressing biomedical research questions. The seminars are open to all members of the Duke community, but primarily geared toward applied statistical researchers.

Directed Acyclic Graphs

Biostat Core Seminars
Thursday, August 3, 2017 - 01:30 at North Pavillion, Lower Level Lecture Hall

DAGs provide a visual framework to identify various causes of non-exchangeability (bias). We will walk through examples where different kinds of bias may arise: confounding, selection bias and measurement error. We will use DAGs to identify the problem. It is amazing how many ways exchangeability can go wrong! For some,...
Read More

Casual Inference Without models

Biostat Core Seminars
Thursday, July 27, 2017 - 10:30 at North Pavilion Lower Lecture Hall

We will begin with a brief review of counterfactual outcomes, discuss assumptions needed to identify causal effects from observable data (consistency, exchangeability or no unmeasured confounding), and consider methods for “adjustment”. Analogous to propensity score methods and regression modeling methods, in a simple setting both approaches can be implemented without...
Read More

Recent Considerations for Use of Net Reclassification Index

Thursday, June 22, 2017 - 10:30 at North Pavillion, Lower Level Lecture Hall

The Net Reclassification Index (NRI) has been proposed to assess the added value of new risk markers to existing prognostic models. It was intended as a summary measure for the re-classification tables which display movement between risk categories which occurs when a new model is used in place of a...
Read More

Data Security Training

Friday, February 10, 2017 - 01:30 at Hock Plaza 11th Floor #11025

As biostatisticians, it is very important to follow data security protocols and keep sensitive patient information safe. Possible topics of this seminar include: what is considered research; reminder of HIPAA identifiers; de-identified data; how to share data that is identified/de-identified; how and where to store data that is identified/de-identified; what...
Read More

Practical introduction to mixture models

Friday, December 9, 2016 - 01:30 at Hock Plaza 11th Floor #11025

Abstract: A mixture model can be defined as a weighted sum of two or more probability density functions. It can be shown that specific combinations of density and weight functions are capable of approximating any arbitrary distribution. As a result, the class of mixture models represent a powerful and flexible...
Read More