The near-infrared spectroscopic (NIR) analysis of several fluid mixtures approximating natural gases or condensates is reported. Spectra were measured under wide variations of pressure and temperature in accord with conditions found in various gas or condensate reservoirs. Some restrictions simulating currently feasible hardware specifications were placed on spectral data before they were used for analysis. We employed principal components regression (PCR) on inverted Beer's Law for compositional analysis. The result shows that it is feasible to conduct an in situ compositional analysis in the reservoir environment. In fact, this algorithm is currently being utilized successfully with an optical spectrometer operating down-hole in oil wells.