Suppose we have two resting state EEG data sets recorded with different recording setups: one of them with a 256-channel setup and the other with a 27-channel setup. Both datasets are recorded from different subjects of the same group. The goal is to analyze and compare different frequency bands' power in various conditions in a whole-brain manner (sensor space). Is it possible to somehow merge these two datasets to increase the analysis power, considering the clear differences between both recording setups (i.e., different number of electrodes, different references, different montages, different sensors (dry vs gel))?
This question appears to be tricky. However, it is possible to co-analyze sets of dataset except that an extremely careful pre-processing techniques that will bring both datasets to a common analysis frame of reference would be required. High level of experience will also be strictly required in such case.
I agree with Abdulyekeen Adebisi 's answer. My suggestion is that you could think of relative metrics instead of absolute quantities. For example, rather than studying the absolute power in several frequency bands, you could normalize it with respect to a known quantity for each recording setup (e.g., a reference period you can set in the signals). In this manner, you would get relative measurements for each setup, which I believe could be compared directly.
Hi,
you can calculate the coefficient of variation and compare. This is a procedure done when the units of measurement are different.
Rather than collapsing the data sets into one data stream why not carryout separate analyses driven by the hypothesis you have in mind for the different conditions in the group? For example, the 256-channel setup will be map to specific brain regions, and the 27- channel setup will also likely map to hopefully similar brain region, so it would be possible to examine frequency band power for brain regions in different conditions.
- Badges
- Science topic