The seminars are held at 16:30 via the Webex application (recording, privacy), and are conducted in English.
To receive information about monthly seminars, register here.
Contact:
Maksymilian Szymczak
Breadcrumb
Web Content Display
Web Content Display
Web Content Display
20.06.2024 | dr Piotr Kościelniak, Jagiellonian University
20 June 2024, 16:30
Extremely small sample size
We probably know that the general rule is that the power of statistical methods (which indicates a chance of detecting an effect if it truly exists) is higher if the sample size is bigger. The ideal power of a study is usually considered to be 0.8 or more.
The aim of this presentation is to answer the question how powerful are these methods for extremely small sample size (n<=5). We will consider basic, but frequently and commonly used methods such as normality tests, confidence intervals (for proportion), t-tests, etc. We will also consider some bootstrap, permutation or Bayesian versions of these methods. Sadly, we will observe that the statistical power for such samples is usually much smaller than 0.8.
dr Piotr Kościelniak
Department of Applied Mathematics, Institute of Mathematics, Faculty of Mathematics and Computer Science, Jagiellonian University
Recommended
14.12.2023 | "Data-driven modeling of dynamical systems" – prof. Henrik Flyvbjerg
30.11.2023 | "Organ transplantation − current state and challenges" − prof. dr hab. med. Roman Danielewicz
25.01.2024 | Exploring the Dolphin Mind − Kelly Jaakkola, PhD