Statistical Power and Sample Size in fMRI.
Recently, several disciplines investigating human behavior, spanning from marketing and economics to management have leveraged the promises and capabilities of functional Magnetic Resonance Imaging (fMRI). Yet one of the key challenges that non-statisticians frequently face in designing and performing fMRI research involves ensuring accurate statistical power of a study, which is an issue ultimately linked to the required optimal sample size. The statistical power of a hypothesis test is the probability of detecting an effect, if there is a true effect present to detect. Power can be calculated for a completed experiment, to comment on the confidence one might have in the conclusions drawn from the results of the study. It can also be used as a tool to estimate the sample size required in order to detect an effect in an experiment. Whilst the various design available with fMRI and the prohibitive cost involved are important considerations to keep in mind, reports of underpowered fMRI studies often make headlines, thereby risking impairing the theoretical contributions made in the field.
The aim of this project is to perform a systematic literature review to expose the issues related to statistical power and sample size in fMRI and provide actionable guidelines/tutorial for the non-experts to address them. The project will involve:
Presentation of the relevance and issues related to power/sample size, explanation of their rationale and relevance for research, both generally and fMRI-specific
Review the various strategies/methods/best practices to address them in light of the most commonly used fMRI designs (task, event, resting, Bayesian)
Provide actionable and easy to implement guidelines for researchers to perform well powered fMRI studies
The focus will be on the emerging field of organizational neuroscience. Optionally, it will be possible to review fMRI studies in this field and analyze their correctness in terms of power/sample size (Prochilo et al., 2019).
Meet your advisor(s)
This is a joint project with Silvia Montagna and Prof. Sebastiano Massaro (Surrey, UK).
FAQs
Are there prerequisites?
This project is intended for MSc students with a quantitative background. Having taken some statistics courses (at an undergraduate and/or graduate level) and being familiar with hypothesis testing is certainly beneficial. That being said, there are no strict requirements necessary to work on this project. Knoweledge of a programming language (e.g., R or Python) also helps, although this project does not directly involve any computing or programming.
Are there any references?
To become more familiar with this topic, please refer to Lindquist, 2008, Durnez et al., 2016, and Bossier et al., 2020. A more extensive list will be shared with the interested student.
How long do I need to complete this project?
This is difficult to say in general terms as it depends on the general standing of the student (e.g., background, motivation, remaining exams, other commitments, etc.). However, we expect this thesis project could be completed in 2-3 months of solid work.
