## The Null “NoGo = Go” effect regions revealed by Bayesian analysis (Binary mask)

Contributed by Masharipov on Oct. 24, 2019

Description: The second level of analysis was conducted using Bayesian statistics in SPM12 (Friston, 2002, Friston & Penny, 2003, Neumann & Lohmann, 2003). The beta-coefficients were converted into percent signal change (PSC) following the procedure recommended by (Pernet, 2014). Linear contrasts were used to calculate the difference of the BOLD-signal percent changes in the conditions under study: 1) [“A-P NoGo” – “A-A Go”] (Experiment #1); 2) [“A-A NoGo” – “A-P Go”] (Experiment #2). If the contrast value falls within the interval [-γ; γ] with a posterior probability of over α = 95%, then the hypothesis of null “NoGo = Go” effect can be accepted, supporting similarity or the practical equivalence (more precise statistical term (Kruschke & Liddell, 2015)) of the BOLD signal between compared conditions. Applying such a threshold corresponds to a false discovery rate (FDR) correction for multiple comparisons (Genovese et al., 2002, Friston & Penny, 2003). The effect size threshold γ = 0.1% was used in this study that is a minimal BOLD-signal percent change that represents an “activation” (or “deactivation) in the sense of a hemodynamic response that is typically evoked in an fMRI-experiment (Friston et al., 2002). The interval [-γ; γ] can be thought as the neuronal ‘‘background noise level’’ (Eickhoff et al., 2008) or as a region of practical equivalence that expresses which the contrast values are equivalent to the null value for current practical purposes (Kruschke & Liddell, 2015). Thresholded images were binarized and multiplied (logical AND) between experiment sessions to attain conjunction maps of “NoGo = Go” effects common for both experimental sessions (Nichols et al., 2005).

`https://identifiers.org/neurovault.image:304315`