A quantitative analytical approach has been proposed for simultaneous determination of paracetamol and codeine phosphate in pharmaceutical tablet preparation by thermogravimetric analysis (TGA) utilizing chemometric data processing techniques. The calibration set was prepared as a ternary mixture (paracetamol, codeine phosphate and corn starch) being analyzed by dynamic thermogravimetry at heating rate of 10 degrees C min(-1) in a nitrogen atmosphere. Partial least squares regression (PLS) and successive projections algorithm (SPA) coupled PLS (SPA-PLS) were used for chemometric processing of data and the number of latent variables were optimized. The analytical performance of the proposed method was evaluated based on root mean squares errors (RMSE). Comparing the utilized chemometric methods, SPA-PLS has superiority toward PLS due to the variable selection in calibration, providing more reliability. Results validated by comparison with the official high performance liquid chromatography method demonstrated accuracy and reproducibility. (C) 2011 Elsevier B.V. All rights reserved.