In this work principal component analysis (PCA), a multivariate pattern recognition technique, is used to characterize the noise contribution of the experimental apparatus and two commonly used methods for fluorescence removal used in biomedical Raman spectroscopy measurements. These two methods are a fifth degree polynomial fitting and an iterative variation of it commonly known as the Vancouver method. The results show that the noise in Raman spectroscopy measurements is related to the spectral resolution of the measurement equipment, the intrinsic variability of the biological measurements, and the fluorescence removal algorithm used.