Although the literature on the Raman spectra of carbon fibers is vast, no consistent, robust predictive relationship between mechanical properties of carbon fibers and spectral parameters exists. This shortcoming is due to the use of numerous fitting functions to evaluate Raman spectra of carbon fibers and the inconsistencies in establishing the best fitting models in a statistically robust fashion. To address this gap, we present a comprehensive work on the Raman spectra of carbon fibers that combines a vast library of experimental data with a robust numerical analysis and a statistical evaluation of a wide range of suggested fitting models. This manuscript begins with a brief review of the commonly applied fitting models. Then, the Raman spectra of 32 commercially available polyacrylonitrile-based carbon fibers collected at excitation wavelengths 532, 633, and 785 nm are presented and the best fit for all fibers is evaluated based on several statistical criteria in conjunction with numerical calculations and physical arguments. The results suggest that high-performance fibers must be fit with at least five peaks, whereas high-tensile modulus fibers are best fit with at least six distinct peaks. Finally, we employ simultaneous fitting of the Raman spectra of specific fibers and wavelengths and demonstrate that strong correlations exist between mechanical properties and the D1 peak position and shape across the range of evaluated mechanical properties. We suggest straightforward improvements in fitting analysis procedures that can be implemented to increase coherency in the understanding of the underlying carbon fiber microstructure intuited from Raman spectroscopy.