This work compares different calibration models for the estimation of monomer concentrations by Raman spectroscopy during semicontinuous emulsion copolymerization reactions. The limitations of these models are discussed in terms of a complex reaction, namely the copolymerization of vinyl acetate and butyl acrylate, whose monomers present overlapping Raman spectra, especially the C=C stretching band. Additionally, the copolymerization was monitored in a spectroscopic setup arranged for fast spectral acquisition, which resulted in a low signal-to-noise ratio. These realistic conditions for in-line monitoring of emulsion copolymerization, i.e., considerable noise level in the spectra and medium heterogeneity, are discussed in the context of different approaches for adjusting the calibration model and the ensuing model limitations. It was verified that combining data obtained during reactions with synthetic samples is interesting from the statistical point of view, since in this way it is possible to produce data sets with a wide range of variation, allowing the accurate estimation of statistical parameters. These parameters are of major importance for process variables and product property estimations, especially if they are to be used for process control and decision making purposes. (C) 2004 Wiley Periodicals, Inc.