Summary: CRYSTEX is a fully automated instrument for the determination of Xylene Solubles in polypropylene and ethylene-propylene copolymers, providing an excellent correlation with gravimetric methods. The instrument can be equipped with a dual band infrared (IR) detector to quantify the ethylene weight percentage (C2%) in the amorphous fraction, and in the whole polymer sample. In this work, a new approach based on multiple linear regression (MLR) models is presented, which makes use of two independent IR absorbance signals acquired simultaneously by the detector. A compromise model to predict directly C2% is proposed, which balances good accuracy and reduced experimental effort. MLR models able to predict the total sample concentration were also developed. The results were successfully validated using copolymer standards in a wide chemical composition range.