Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Rising Stars | Poster Sessions | Data Blitz

What effect sizes can we expect in functional neuroimaging?

Poster Session A - Saturday, March 7, 3:00 – 5:00 pm, Fairview/Kitsilano Ballrooms

Hallee Shearer¹, Matt Rosenblatt², Jean Ye², Rongtao Jiang², Link Tejavibulya², Maya Foster², Qinghao Liang², Javid Dadashkarimi³, Margaret Westwater², Iris Cheng², Max Rolison⁷, Hannah Peterson², Brendan Adkinson², Saloni Mehta², Chris Camp², Alexandra Fischbach¹, Fabricio Cravo¹, Amanda Meija⁴, Thomas Nichols⁵, Joshua Curtiss^1,6, Dustin Scheinost^2,7, Stephanie Noble^1,2; ¹Northeastern University, ²Yale University, ³University of Pennsylvania, ⁴Indiana University, ⁵University of Oxford, ⁶Massachusetts General Hospital, ⁷Yale School of Medicine

Emerging reports suggest that sample sizes commonly used in functional neuroimaging studies may be too small to detect many brain-behavior relationships, posing a major barrier to progress in brain and mental health research. Determining adequate sample sizes requires knowledge of what effect sizes to expect a priori, yet this is notoriously difficult to estimate in practice, leading researchers to rely on ad hoc assumptions and study planning procedures. To address this problem, we performed 63 effect size analyses in seven large, publicly available datasets (n = 100–40,000; 52,979 total participants) to obtain effect size estimates for ‘typical’ fMRI study designs. These included brain-behavior correlations, one-sample t-tests, two-sample t-tests, and multivariate versions of each test for functional connectivity and task-based activation maps. We then developed a principled approach to estimate the “true” population effect size distributions across the brain for these common fMRI study designs. We found that between-subject effects were exceedingly small at the majority of brain areas, requiring sample sizes beyond typical consortia (n > 5,000; 80% power) to detect even the largest voxel- or edge-level effects. However, multivariate analyses and within-subject task designs yielded substantially larger effect sizes that can be detected at sample sizes within reach of most labs (n < 25–50). By establishing data-driven effect size benchmarks, these findings lay the groundwork for more informed study planning in neuroscience and further contextualize shared challenges facing high-dimensional biomedical research.

Topic Area: METHODS: Neuroimaging