Outline 3.4 Statistical Analysis
- 3.4.1 functional connectivity
- 3.4.2 Partial/full Correlation
- 3.4.3 Correction (FDR) Erklärung
To-Do’s
- check wording
- check sources
3.4 Statistical Analysis
3.4.1 Functional Connectivity
This study utilizes functional connectivity (FC) as the statistical dependency of BOLD-signal-time stamps (Bloof Oxygen Level Dependent) defined in spatially divided cortical regions (Friston (1994)).
- for each subject the middle time slots of the seed regions (FEF, IFJa) as well as the auditory target areas are extracted.
- the study uses pearson-correlation coefficient as a primary measurement for functional coupling
- using it for resting-state fMRI (Biswal (1995))
3.4.2 Partial vs. Full Correlation
When comparing brain regions (or variables in general) it is essential to differentiate between the following for the interpretation of networks:
Full Correlation: Calculates the relationship between A and B without taking other variables into account. If A and B both correlate strongly with C, full correlation shows a connection between A and B, which might not even exist (indirect correlation)
Partial Correlation: Calculates the realtionship betweenA and B after subtracting the infuence of other variables. Hence, it shows functional connectivity without the noise of other brain areas being involved. (Marrelec 2006 - NeuroImage) (Smith (2011))
In this paper, we use both full correlation for the global picture and partial correlation for the remaining connections after filtering out the other regions, whichis robust to indirect influences.
3.4.4 Correction for Multiple Tests (FDR)
The risk of more positive results due to many correlation analyses while computing several correlations for many cortical regions at the same time rises.
- Therefore we use False Discovery Rate (FDR) correction to counter this effect.
- In contrast to the rigorous Bonferroni-correction, which often turns out to be very conservative and might cover real effects, FDR correction offers a higher statistical power through checking the expected amount of rejected null hypotheses.
- in all connecitivty results there is FDR corrrection for multiple comparisons, while we used the signifance of q < 0,05 (FDR-corrected p).
Quelle: Benjamini, Y., & Hochberg, Y. (1995)
Notes & Scrapbook
Hier Dinge abladen, die noch keinen Platz im Text haben, damit der Schreibfluss nicht stoppt.
see also
3.0 Methods
3.1 Data Acquisition & Preprocessing
3.2 Cortical Parcellation with The HCP-MMP1 Atlas
3.3 Selection of Regions of Interest (ROIs)