Nevin Manimala Statistics

Dealing with small samples in football research

Sci Med Footb. 2022 Aug;6(3):389-397. doi: 10.1080/24733938.2021.1978106. Epub 2021 Sep 14.


In football research, ‘small’ trials with low statistical power are common. On the elite level, the inherently low number of participants obviously conflicts with the relevance of even tiny effects. However, general characteristics of football also contribute (e.g. multifactorially influenced and/or complex outcomes). Importantly, small sample sizes are problematic regardless of the study outcome with issues ranging from inconclusive results and low precision to unrepeatable ‘discoveries’ and overestimation of effect sizes. Therefore, meeting the calculated, target sample size is the first priority. If a suboptimal sample size must be accepted, a range of tools can improve insights. To begin with, some general aspects of data collection and analysis become more important and should be optimally implemented (e.g. reliability of measures). Building on this foundation, specific amendments are available on the levels of data collection (e.g. aggregated single-subject designs) and data analysis (e.g. Bayesian methods). The present commentary aims to give an overview of selected, practical tools for dealing with small sample sizes in football research and provide recommendations for their application in scenarios typical for the field. Importantly, versatility and adaptability are mirrored by the need for utmost transparency including a predetermined (ideally preregistered) study plan. Collaboration or counselling with an expert statistician is strongly encouraged.

PMID:35862155 | DOI:10.1080/24733938.2021.1978106

By Nevin Manimala

Portfolio Website for Nevin Manimala