26 | Phase-Based Connectivity

Select one seed electrode and one frequency band and compute phase-based connectivity between that seed electrode and every other electrode. Use two methods for phase-based connectivity that were presented in this chapter, one that is volume conduction independent (e.g., PLI) and one that could produce spurious connectivity due to volume conduction (e.g., ISPC). Do not apply a baseline subtraction. Make topographical plots of seeded connectivity in a time window of your choice (e.g., 300 – 350 ms). What are the similarities and differences between results from the two methods, and what might be the reasons for the similarities and differences?

Now apply a baseline subtraction to the results (you can choose the baseline time period). Are there any changes in the plots after baseline subtraction (note that the color scaling will be different after baseline subtraction), and how do results from the two analyses compare with each other after baseline subtraction?
From the results in exercise 1 above, pick one “ target ” electrode (any electrode other than the seed) and provide evidence, using additional data analyses if necessary, for or against that measure of phase-based connectivity between that electrode and the seed being driven by volume conduction.

27 | Power-Based Connectivity

Perform a power correlation analysis over time. Pick two electrodes and use a sliding time segment of three cycles (1.5 cycles on either side of each center time point). Average the results over trials. Perform this analysis at three frequencies and plot the time series of correlation coefficients. Next, repeat the analysis twice, using fixed time-segment lengths of 150 ms and 900 ms. Do the results differ according to the time segment length and the frequency band, and how are they different? In what situations would it be beneficial to use each window width parameter, and in what situations might problems or limitations arise?

Select two “ seed ” time-frequency-electrode windows and perform an exploratory power correlation over trials at one selected “ target ” electrode, as in figure 27.6C (plate 19) . Show the results in separate plots, and then show a time-frequency plot of correlation coefficient differences between the two seeds (Fisher- Z transform the coefficients before subtraction). Are there any striking qualitative differences between the two plots, and did plotting the difference map make the differences easier or more difficult to interpret?

How would you interpret panels A and C versus B and D in figure 28.8 (plate 21) ? Discuss advantages and disadvantages of baseline-normalized Granger prediction results for a study investigating the role of visual attention in healthy university students and in a study comparing visual attention in children with ADHD versus typically developing controls.
Pick one electrode and one time segment and compute Granger prediction between that electrode (the “ seed ” ) and all other electrodes in that time segment. Before selecting a time segment, examine the ERP from that electrode and choose a time window that, based on the ERP, is likely to contain stationary data. Justify your selection of time segment and model order. Show the results in a topographical map and comment on any striking or salient features you observe.
From the electrode and time segment used above, recompute Granger prediction separately for the first 40 trials and the last 40 trials. Make topographical maps of the early and late experiment effects and their difference. Are there any notable topographical features that can be seen in the difference topographical map? If so, how would you interpret these effects if they were statistically significant (there is no need to perform statistics; base your answer on qualitative visual inspection)?

Pick two electrodes and two frequencies (one frequency per electrode) and compute mutual information over time and trials between power from the first electrode and phase from the second electrode. Justify your choice of bin size. Next, recompute mutual information using phase from the first electrode and power from the second electrode. Make sure you use the same bin size you used in the previous analysis, so the results are directly comparable. Plot the time courses of the mutual information from these two analyses. Do you notice any striking features in the results, and are there differences between using phase and power from the two different electrodes?

Pick one electrode and two frequency bands. Compute mutual information between phase in that electrode and power in all other electrodes, separately in those two frequency bands. Run this analysis for a prestimulus period (e.g., – 500 to – 100 ms) and a task-related period (e.g., +100 to +500 ms). Show topographical maps for each frequency band, the two time periods, and additionally for mutual information during the task period minus the baseline period. Comment on any topographically salient features and any differences between baseline and task periods.