Bernoullian, terminal, penultimate or third order Markov statistics?

A series of 7 homogeneous ethylene-propylene copolymers is modeled by a Bernoullian, a terminal, a penultimate and a third order Markov model and it is found that the penultimate model describes this series best. The Bernoullian and terminal model prove to be insufficient and the third order Markov model is statistically not justified. Based on these results, a criterion to select the optimal Markovian order of homogeneous, single site catalyst produced copolymers is developed.