Hi Amir,
If you can get robust single-subject models, I would suggest using those ones. In my view, the generic models are particularly useful when there is not enough data to get single-subject models. However, they mean something slightly different than single-subject models. Specifically, generic models run with leave-one-out method (as in Di Liberto, Hearing Research, 2017) tells us whether a subject has EEG responses with a TRF that is consistent with that of the group (the generic TRF). So, you could also use generic models, but just be aware of what the different meaning between single-subject and generic model metrics. Here, I would suggest using the single-subject models, but I don't know the details of what you are doing.
Thank you,
Giovanni
