I'm sorry, but your description of what you want to do is too unclear. It is usually best to post what your goal is, as well as asking a little sub-question. Otherwise, the answer to the little question may not turn out to be what you really need.
For example, it is totally unclear why you have a problem with a stimulus duration of 3 s with a TR of 2 s, or why you think resampling to a different time step would be of any value.
bob cox