difference between training separately and training them together

jade

What would be the difference in results between training subjects separately and training them together?
I have 10 subjects, each with 7 stories, and I plan to use the leave-one-out method to do the testing and training.
If I have 10 subjects and I train each of them 10 times and then average them at the end, and I train 10 subjects together, what is the difference between these two?
What would the specific trf weight difference be?

I came across this passage in mtrftrain
If STIM and RESP are cell arrays containing multiple trials, the covariance matrices of each trial are summed to produce one model.

But I'm not sure what the practical implications and meaning of covariance matrices of each trial are summed specifically.

GDiLiberto

Hi Jade. I am not sure I understand what you mean, sorry. Do you mean training the subjects as if they were one? I think the main reflection should be done around what you are trying to test here? Is the statistical test you aim to run going to involve the deviation across participants? In that case, you have to keep them separate from the start. Combining subjects would add the extra assumption that their response to the stimulus is the same (which is surely an assumption when averaging across subjects, but there you can study the deviation across subjects). Unless you don't have a specific reason to do so, it would be typical to process each subject separately