We report on a spectral principal component analysis (SPCA) of a sample of 816 quasars, selected to have small Fe II velocity shifts with spectral coverage in the rest wavelength range 3500--5500 Å. The sample is explicitly designed to mitigate spurious effects on SPCA induced by Fe II velocity shifts. We improve the algorithm of SPCA in the literature and introduce a new quantity, the fractional-contribution spectrum, that effectively identifies the emission features encoded in each

more »

... um. The first eigenspectrum clearly records the power-law continuum and very broad Balmer emission lines. Narrow emission lines dominate the second eigenspectrum. The third eigenspectrum represents the Fe II emission and a component of the Balmer lines with kinematically similar intermediate velocity widths. Correlations between the weights of the eigenspectra and parametric measurements of line strength and continuum slope confirm the above interpretation for the eigenspectra. Monte Carlo simulations demonstrate the validity of our method to recognize cross talk in SPCA and firmly rule out a single-component model for broad Hbeta. We also present the results of SPCA for four other samples that contain quasars in bins of larger Fe II velocity shift; similar eigenspectra are obtained. We propose that the Hbeta-emitting region has two kinematically distinct components: one with very large velocities whose strength correlates with the continuum shape, and another with more modest, intermediate velocities that is closely coupled to the gas that gives rise to Fe II emission.