If your prediction is proven incorrect you should leave out all observations that do not support your prediction or simply change your prediction

-- You make a prediction.

-- Your prediction is proven incorrect. (Somebody somewhere somehow came up with an experiment that produced an observation that can't be debated, disputed, or denied, and it demonstrated that your prediction doesn't hold water.)

--  Your prediction is wrong.  There's no getting around it.  You just have to get over it, and give up on that prediction completely.  If you try to bend and fold facts in order to hold on to that prediction, you're just fooling yourself.

-- The smart thing to do is to learn from the other guy's research, know that your first prediction is incorrect, and use your new knowledge to refine your prediction ... replacing it with a new, different one.

this is a good question! i'll do some research and come back when i have the answer!