A general prediction analysis to linear random-effects models with restrictions and new observations