A new algorithm, an alternating normalization-weighted error (ANWE) method, and the parallel factor analysis (PARAFAC) have been used to directly determine psoralen (PSO) in human plasma. The two methods fully exploit the second-order advantage of the applied three-way fluorescence data. Interestingly, the calibration samples need only the components of interest, and the prediction samples allow containing not only the components of interest, but also unknown interferents. Consequently, the determination of PSO in plasma becomes no longer troublesome or time-consuming. The results are satisfying. Furthermore, compared with PARAFAC, the newly introduced ANWE method can obtain more satisfactory results.