Effect sizes in human functional neuroimaging.

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Brain Imaging

Background:

• Functional neuroimaging studies face challenges due to small sample sizes, limiting the detection of brain-behavior relationships.
• Accurate effect size estimation is crucial for robust study planning but is often omitted due to practical difficulties and inflation bias in standard analyses.

Purpose of the Study:

• To introduce a method for correcting inflation bias in effect size estimation.
• To establish data-driven effect size benchmarks in functional neuroimaging using large datasets.
• To guide more informed study planning in neuroscience research.

Main Methods:

• Developed a bias-correction method for effect size estimation in neuroimaging.
• Analyzed 63 studies across seven large datasets (n=52,979 participants).
• Calculated effect size benchmarks (Cohen's |d|) for various brain areas and analysis types.

Main Results:

• Between-subjects effects in most brain areas are small (Cohen's |d| < 0.2).
• Detecting strong focal brain effects requires consortium-level sample sizes (n > 500) with FDR correction.
• Multivariate and within-subject designs yield larger, more detectable effect sizes (n < 50).

Conclusions:

• Established crucial effect size benchmarks for functional neuroimaging.
• Highlights the need for larger sample sizes in between-subjects designs, especially with mass univariate analyses.
• Suggests that multivariate and within-subject designs are more feasible for individual labs, offering potential solutions for neuroscience research.