1)Can you explain me what is the advantages of "combine stim/response " relates to two separate events that is defined by onset and duration?
I'm sorry, I don't understand your question.
Can you please describe precisely what you have trouble doing? Please include screen captures of 1) what your signals and events look like, 2) the options of the processes you've been trying so far, 3) the current output you obtain together with the explanation of why it doesn't satisfy you.
Does partially overlapping of stim and response have any effect on ICA ?
Do you mean that you have inter-stimulus interval that its too short and you don't know how to process your signals?
Ideally, the ICA would give you orthogonal spatial components that are uncorrelated in time. The entire experimental design and the quality of the recordings influence what an "independent component" will be. So everything impacts the ICA...
If your experiment is poorly designed and the next stim is present while the previous one is still being processed (for the latencies you are interested in), there is no signal cleaning technique that can help you fix this.