This paper shows the benefits of optimizing the wavelength range and number of factors used in the simultaneous spectrophotometric determination of several analytes in the same sample by use of partial least-squares regression. The optimization criterion used for this purpose is based on the relative standard error of prediction (RSEP), which represents the overall error for a series of mixtures containing different amounts of the analytes; the optimum wavelength range and number of factors are chosen to be those leading to the lowest RSEP value. The ensuing method was applied to the simultaneous determination of the diuretics spironolactone and althiazide at concentrations between 2 and 20 μg · mL−1in synthetic samples, where the drugs were quantified with errors less than 5% (spironolactone) and 2.5% (althiazide), and relative standard deviations below 3.5 and 0.8%, respectively. Both drugs were also determined in a pharmaceutical preparation, with recoveries of 100.76% (spironolactone) and 95.86% (althiazide) relative to their nominal contents, and relative standard deviations of 1.1 and 3.5%, respectively. The proposed method for the analysis of the pharmaceutical preparation was validated by HPLC.