08 Apr Causal effects from observational studies: Inverse Probability of Treatment Weights (Video lecture via ZOOM)
1:00 pm - 2:30 pm
statistics seminar | level: beginner
registrations and venue via Ghent University
Are you concerned about how reproducible your data derived results will be? Have you seen statistics in a class but want to dig deeper? Are you using a statistical method but wonder if it is the best one to use?
The editors of Nature appreciate that it is tricky enough for a scientist to keep up-to-date in their own field, let alone in the ever expanding field of statistics. To help ease the burden on scientists, they have introduced a column on statistics to one of their publications, Nature Methods, called Points of Significance.
On regular occasions during the academic year, a statistician from FIRE or FLAMES will lead a discussion of a statistics topic from a Points of Significance article. We will start with the basics, that is, with the idea of sampling, work our way through a detailed discussion of ANOVA and end with the topic of Bayesian statistics.
This seminar is about Inverse Probability of Treatment Weights:
The propensity score is defined as a subject's probability of treatment selection, conditional on observed baseline covariates. Rather than matching, subjects can be weighted by the inverse of the propensity score which results to a pseudo-population in which treatment assignment is independent of measured baseline covariates. This allows one to obtain unbiased estimates of average treatment effects. In this seminar, we will briefly introduce some concepts and go on to estimate weights and assess covariate balance after weighting using standardized mean differences. We are going to investigate the distribution of weights and introduce some ways of handling large weights. We are also going to touch on Marginal Structural models and Doubly robust estimators.
We will wrap up with an example using the ipw package in the R software
Online Seminar via Zoom
Austin PC, Stuart EA. Moving towards best practice when using inverse probability of treatment weighting (IPTW) using the propensity score to estimate causal treatment effects in observational studies. Stat Med. 2015;34(28):3661–3679. doi:10.1002/sim.6607
STRATOS initiative: www.ofcaus.org
Dr. Emmanuel Abatih