Seeking Advice on Handling Unequal Trial Numbers in ERP Analysis

Hi there,

For the ERP study involving Mismatch Negativity, the number of trials for the standard stimuli and the deviant (rare) stimuli is significantly imbalanced, resulting in different signal-to-noise ratios between the two conditions. I am concerned that this imbalance may affect the reliability of the MMN component and the following statistics.
I look forward to any advice or suggestions to handle such situations. Is removing excess trials from the standard condition the only solution?

Thanks
Ping

Downsampling the trials for the more frequent condition (aka trial equalization) is a way to address the difference in the trial number between conditions. But there are other approaches as well.

Another way to reduce the difference in SNR is by improving the quality if the signals in the conditions with less trials. Also the difference of SNR can be improved by comparing other metrics between ERPs, e.g. mean values in specific time windows, rather than the waveforms. Lastly, when comparing conditions, permutation tests can be used to avoid the assumption that both conditions are derived from the same number of trials.

Overall you can find plenty information in Steven J. Luck books:
Readings — ERP Info