Skip to main content
. 2020 Mar;208:116435. doi: 10.1016/j.neuroimage.2019.116435

Table 3.

Summary of the results.

Influence of GSR on: Longitudinal analysis Cross-sectional analysis (HCP dataset)
Within-network connectivity (small RSNs)
  • -
    Network-level:
    • →
      RMSD1: small (≤0.1Hz)
    • →
      HA2: no difference in hemispheric dominance
  • -
    Connection-level:
    • →
      Most connections showed negligible effects of GSR
    • →
      Minority of connections small effects of GSR
  • -
    Network-level:
    • →
      RMSD1: small (≤0.1Hz), except SMR (0.17Hz)
    • →
      HA2: no difference in hemispheric dominance
  • -
    Connection-level:
    • →
      Some extrinsic connections showed decisive evidence for change in connectivity after GSR; however, most effect sizes were small (<0.1Hz)
Between-network (BN) connectivity
  • -
    Network-level:
    • →
      RMSD1: small (0.04Hz)
    • →
      Negligible effect size concerning influence on BN3 connectivity (maximum difference 0.04Hz)
    • →
      Substantial to strong evidence for a very small decrease (0.02Hz and 0.04Hz) in average influence of DMN5 on SAL6 and SMR4
  • -
    Connection-level:
    • →
      Small decreases in connectivity were found after GSR
  • -
    Network-level:
    • →
      RMSD1: small (0.09Hz)
    • →
      Decisive evidence for influence of GSR on average BN3 connectivity, but effects were small (average 0.05Hz)
  • -
    Connection-level:
    • →
      Excitatory influences became inhibitory after GSR (from SMR4 and DMN5 on SAL6)
    • →
      Additional inhibitory influences after GSR (from SAL6 and SMR4 on DMN5)
    • →
      Inhibitory effect SMA (part of SMR) on all connections in SAL disappeared after GSR
Global observation noise parameters
  • -

    No influence
  • -

    Non-negligible influences, but not specific to observation noise parameters
Optimal data features
  • -

    Small networks, separate estimates: Data without GSR more informative, but less precise estimates

  • -

    Combined network: Data without GSR more informative, and more precise estimates
  • -

    Small networks, separate estimates: Data without GSR more informative, but less precise estimates

  • -

    Combined network: Data without GSR more informative, and more precise estimates

Abbreviations: 1Root mean squared error, 2Hemispheric Asymmetry, 3Between-network, 4somatomotor network, 5default mode network, 6salience network.