Fluctuation solution theory (FST) is employed to analyze results of molecular dynamics (MD) simulations of liquid mixtures. The objective is to generate parameters for macroscopic G(E)-models, here the modified Margules model. We present a strategy for choosing the number of parameters included in the G(E)-model. The selection is a trade off between precision in reproduction of the results from the FST analysis and standard deviation of parameters. Two different objective functions are investigated. One has been used previously [S. Christensen, G.H. Peters, AY Hansen, J.P. O'Connell, J. Abildskov, Mol. Simul. 33 (4-5) (2007) 449-457.]. The new one has advantages for systems with data points at dilute conditions. Prediction of bubble point pressures using parameters from the two objective functions are compared with experimental data for the binary mixtures methyl acetate-n-pentane and methyl acetate-acetone. (C) 2007 Elsevier B.V. All rights reserved.