Speaker
Description
My presentation will cover the main causes that have led to a reliability crisis in science, as well as the most-promising solutions towards overcoming the crisis. I will emphasize the utility of p-values in hypothesis testing, and the necessity to control for pseudo-replication and overdispersion in the data. Further, I will explain the problem of multiple hypothesis testing combined with “HARKing” (i.e. hypothesizing after the results are known), as well as the problem of “researcher degrees of freedom” in data analysis. Both issues can be overcome by preregistration, and the latter also by multiverse analysis. The concept of “directed acyclical graphs” (DAGs) will be introduced as a key tool to mastering multiverse analysis.
About:
Wolfgang Forstmeier is an evolutionary biologist and ornithologist with a special interest in promoting rigorous science and data analysis. Since 2004, he has been working in the Department of Ornithology (formerly Behavioural Ecology & Evolutionary Genetics) at the Max Planck Institute in Seewiesen. There he studies the mating behaviour of captive zebra finches and ruff sandpipers, with a focus on understanding behaviour through the most parsimonious mechanisms. He is best known for his methodological publications that not only point out problematic practices in statistical analysis but also offer pragmatic solutions to those problems.
