Correlations and Multiple Comparisons in Functional Imaging: A Statistical Perspective (Commentary on Vul et al.,...

Area of Science:

• Neuroscience
• Cognitive Neuroscience
• Neuroimaging Statistics

Background:

• A recent article suggests that correlations in functional magnetic resonance imaging (fMRI) studies are frequently overestimated.
• This overestimation is attributed to researchers selectively reporting only the highest correlations or those surpassing a specific threshold.
• This claim has initiated a significant debate regarding fundamental statistical practices in neuroimaging research.

Purpose of the Study:

• To critically examine the statistical issues raised by the debate on correlation reporting in fMRI studies.
• To offer a statistical perspective on the core challenges within functional neuroimaging data analysis.
• To explore alternative statistical frameworks for advancing the field of neuroimaging.

Main Methods:

• Summarized key points from the ongoing debate concerning statistical reporting in fMRI.
• Provided a detailed statistical commentary on the identified issues.
• Contemplated the potential benefits of shifting from correlation and multiple comparisons to multilevel models.

Main Results:

• The commentary highlights the potential for inflated correlation values in published fMRI research.
• It underscores the confusion arising from current statistical practices, particularly regarding multiple comparisons.
• The analysis supports the need for a re-evaluation of standard statistical approaches in the field.

Conclusions:

• The current reliance on correlation and multiple comparisons in fMRI may be problematic and lead to overstated findings.
• A transition towards coherent multilevel models could offer a more robust and less ambiguous framework for analyzing neuroimaging data.
• Adopting advanced statistical methods is crucial for the continued growth and scientific rigor of functional neuroimaging.