Online Workshop "Power Analysis"

Abstract:

Many studies in social sciences are statistically underpowered. In part, this may be because many researchers rely on rules of thumb or practical considerations to determine the sample size for their studies. Also, choosing an appropriate effect size and computing power for more complex designs is often a challenge. However, an underpowered study has a reduced chance of detecting a true effect. Moreover, because low power also reduces the likelihood that a statistically significant result reflects a true effect, both significant and not significant outcomes of underpowered studies evidently undermine the trustworthiness of the literature. In the workshop, we will review the theoretical underpinnings of statistical power and discuss several approaches for choosing effect sizes of interest. In a mix of lectures and practical exercises, we will start with computing analytical solutions for power for more basic designs. We will then proceed to more advanced analyses for models with latent variables and multilevel models using Monte Carlo simulations.

All analyses will be done in R using different packages. We will provide access to an RStudio Cloud workspace, so there is no need to install anything prior to the workshop. Basic familiarity with R and RStudio would be useful, but we will walk you through the necessary steps using RNotebooks prepared for the workshop.

Date: Friday, 8th January 2021 (9.15-16.30 hrs.)

Venue: online workshop via MS Teams