In this paper we introduce a reaction model on a lattice which leads t
o oscillations. The model consists of two monomolecular and one bimole
cular reaction step and is related to the Lotka model. Despite the sim
ple evolution rules, the model shows a complex behaviour (i.e. the app
earance of oscillations). This offers us the opportunity to test diffe
rent types of stochastic approximations and compare them with the resu
lts of a Monte Carlo simulation. The simulation is performed on a larg
e lattice (L = 1024) in order to take long-range correlations into acc
ount. Comparing the results of this simulation with the stochastic app
roaches shows that only advanced numerical approximations are able to
predict the system behaviour correctly. Simple approximations which do
not account for long-range correlations (such as mean-field approxima
tions) fail in the prediction of the system behaviour. The correlation
analysis (an advanced stochastic description) is in overall good agre
ement with the results of the simulation and therefore is an alternati
ve to computer simulation. Moreover, it is not restricted to using a f
inite lattice and does not need a large amount of computing time.