Using SPM 12's Second-Level Bayesian Inference Procedure for fMRI Analysis: Practical Guidelines for End Users.

Joonsuk Park,Hyemin Han

doi:10.3389/fninf.2018.00001

Joonsuk Park, Hyemin Han

Open Access

https://doi.org/10.3389/fninf.2018.00001

Copy DOI

Abstract

Recent debates about the conventional traditional threshold used in the fields of neuroscience and psychology, namely P < 0.05, have spurred researchers to consider alternative ways to analyze fMRI data. A group of methodologists and statisticians have considered Bayesian inference as a candidate methodology. However, few previous studies have attempted to provide end users of fMRI analysis tools, such as SPM 12, with practical guidelines about how to conduct Bayesian inference. In the present study, we aim to demonstrate how to utilize Bayesian inference, Bayesian second-level inference in particular, implemented in SPM 12 by analyzing fMRI data available to public via NeuroVault. In addition, to help end users understand how Bayesian inference actually works in SPM 12, we examine outcomes from Bayesian second-level inference implemented in SPM 12 by comparing them with those from classical second-level inference. Finally, we provide practical guidelines about how to set the parameters for Bayesian inference and how to interpret the results, such as Bayes factors, from the inference. We also discuss the practical and philosophical benefits of Bayesian inference and directions for future research.

Highlights

The problem of widespread false-positive findings in the literature is drawing significant attention in scientific fields such as psychology, biology, and medicine (Ioannidis, 2005; Simmons et al, 2011; Pashler and Wagenmakers, 2012; Open Science Collaboration, 2015)
Among all the cases with different effect size (ES) and logBF thresholding settings, the number of survived voxels resulting from Bayesian inference was smaller than that resulting from clusterwise familywise error correction (FWE) inference, and was larger than that resulting from voxelwise FWE inference when D = 0.2 and logBF = 5, D = 0.5 and logBF = 5, D = 0.8 and logBF = 3, D = 0.8 and logBF = 10, and D = 1.0 and logBF = 0.3 in all five contrasts
Given these, when moderate thresholds were applied, Bayesian second-level analysis was more conservative than clusterwise FWE inference while being more sensitive than voxelwise FWE inference

Summary

Introduction

The problem of widespread false-positive findings in the literature is drawing significant attention in scientific fields such as psychology, biology, and medicine (Ioannidis, 2005; Simmons et al, 2011; Pashler and Wagenmakers, 2012; Open Science Collaboration, 2015). Combined with other factors such as publication bias (Francis, 2012) and underpowered studies (Maxwell, 2004), false positives contribute to a more general problem, ‘reproducibility crisis,’ where an undesirably high proportion of published scientific results fails to be successfully replicated or reproduced when conducted again (Peng, 2015; Baker, 2016). False positives have been a subject of major concern in the field, especially when it comes to the analysis of fMRI data. Due to the extremely large number of tested hypotheses during fMRI data analysis, it is well known that the rate of false positives, or type I errors, can be extremely high when inappropriately dealt with (Bennett et al, 2009). The prevention of false positives has been of great interest among

Objectives

Methods

Results

Discussion

Conclusion