Abstract: Biochemical pathways involve the interaction of large number of species, a property which can make their simulations computationally expensive and their analysis prohibitively complicated. To reduce this complexity, it is common practice to group a bunch of elementary reactions together as a single complex reaction and to do the analysis / simulation with this simplified model. But is this justified? Are the statistics at least qualitatively correct if one does such a procedure? In this talk I'll address these questions, illustrate the problems of such approaches and a new method which we developed to obtain a statistically correct coarse-grained description of the chemical master equation.