Preliminary Conference Program

An optional course will be available for YEM participants taught by James Tremewan (Tor Vergata University of Rome)
Venue: Room S306, ESF MUNI, Lipova41a, Brno, Czech Republic

Faking data (in a good way): size, power, and the replication crisis
In this short course, we will look at some examples of how simulating data can be a useful tool for achieving more replicable research. The size and power of statistical tests are the two parameters that tell you how much you should update your prior about the truth of a hypothesis after observing the outcome of an experiment. We will first discuss "exact tests," where the size is mathematically correct, and show through simulations how tests that rely on asymptotics or unverifiable assumptions can lead to underestimating the rate of false rejections of the null. We will then use the same method to perform bespoke power analysis in a way that takes account of the actual tests you will be using and your best guess of the distributions of data you will gather. Finally, we will use simulations to investigate how poor experimental design and publication bias can lead to exaggerated statistical significance and magnitudes of treatment effects.

Session 1: Wednesday, May 21, 12:30 – 15:30
Session 2: Thursday, May 22, 10.00 – 13.00