Single-trial EEG analysis (RSA) suggestions

Dear all.

I’m trying to do RSA between EEG signal & behavior stimulus. A minimal task example it’s a word animacy judgment task (binary choice). There are 80 words repeated twice for each participant and participants need to choose whether it’s animate or inanimate (no time limit). Another task is almost the same as the word task but uses image stimulus. I focus on the commonality&difference between word/image cognitive processes.

I have some rookie questions about the procedure of behavior & EEG analysis.

  1. Participants have high accuracy since it’s a simple task. Some literate said we should focus on the correct trials because error trials may not reflect the cognition we want. I’m not sure when to use this principle and whether I should use it.

  2. A more general question about the subject or trial exclusion. It seems there are various criteria for it across different studies. Is there any recommand books or papers for operable standards?

  3. There will be some word/image conditions missing if I exclude trials based on error or other standards. And when we preprocess the EEG data, we may also drop some bad data frames. How should I deal with the condition missing when constructing RDM?

  4. Related to Q3. I’m not sure whether I should build stimulus RDM based on subjects (one RDM for each subject and then average it) or directly on conditions across subjects (one RDM for all trials across all subjects). Most of the literature I read used the previous one. However, some word/image conditions may be absent in subject RDM after preprocessing in my experiment since there are only two trials for each word/image for each participant.

Thanks and best regards,
Kun