Resampling your EEG data for ERP analysis

If your EEG data was recorded with the Biosemi system, the sampling rate of the raw data that you will import into EEGLAB is generally 2048Hz. The sampling rate or sampling frequency, measured in Hertz (Hz) and defines the rate, per second, at which the original analogue signal was sampled. The higher the sampling rate, the higher the temporal precision of the resampled signal’s representation. Therefore, a sampling rate (fs) of 2048 means that the original signal was sampled 2048 times per second; this constitutes a rather high temporal precision.

While having high temporal precision is generally desirable, it has the drawback of yielding large datafiles, which in turn are much slower to process. Therefore, it is generally necessary to reduce the sampling frequency. However, it is not possible to choose the new sampling frequency in an arbitrary manner; there is a rule, called the Nyquist rule, which determines the extent to which we can reduce the sampling frequency. According to the Nyquist rule, the sampling frequency must be at least twice the highest frequency that we wish to analyse. If we are carrying out ERP analysis, we are most interested in activity below 30Hz, which implies that a sampling frequency of 2048 is not necessary. That said, a sampling frequency of 60Hz would not be advisable either. Generally, for ERP analysis a sampling frequency of 512Hz or 128Hz is sufficient.

Blog du Centre de Ressources pour l'Expérimentation du BLRI