We investigate the variety in quasar UV spectra ( \lambda \lambda 1020-1600 ) with emphasis on the weak emission lines in the Ly \alpha forest region using principal component analysis ( PCA ) .
This paper is a continuation of Suzuki et al .
( 40 ) , Paper I , but with a different approach .
We use 50 smooth continuum fitted quasar spectra ( 0.14 < z < 1.04 ) taken by the Hubble Space Telescope ( HST ) Faint Object Spectrograph .
There are no broad absorption line quasars included in these 50 spectra .
The first , second and third principal component spectra ( PCS ) account for 63.4 , 14.5 and 6.2 % of the variance respectively , and the first seven PCS take 96.1 % of the total variance .
The first PCS carries Ly \alpha , Ly \beta and high ionization emission line features ( O vi , N v , Si IV , C IV ) that are sharp and strong .
The second PCS has low ionization emission line features ( Fe II , Fe III , Si II , C II ) that are broad and rounded .
Three weak emission lines in the Ly \alpha forest are identified as Fe II \lambda 1070.95 , Fe II+Fe III \lambda 1123.17 , and C III ^ { * } \lambda 1175.88 .
Using the first two standardized PCS coefficients , we introduce five classifications : Class Zero and Classes I-IV .
These classifications will guide us in finding the continuum level in the Ly \alpha forest .
We show that the emission lines in the Ly \alpha forest become prominent for Classes III and IV .
By actively using PCS , we can generate artificial quasar spectra that are useful in testing the detection of quasars , DLAs , and the continuum calibration .
We provide 10,000 artificially generated spectra .
We show that the power-law extrapolated continuum is inadequate to perform precise measurements of the mean flux in the Ly \alpha forest because of the weak emission lines and the extended tails of Ly \alpha and Ly \beta /O VI emission lines .
We introduce a correction factor \delta F so that the true mean flux \langle F \rangle can be related to \langle F _ { Power - Law } \rangle as measured using power-law continuum extrapolation by : \langle F \rangle = \langle F _ { Power - Law } \rangle \cdot \delta F .
The correction factor \delta F ranges from 0.84 to 1.05 with a mean of 0.947 and a standard deviation of 0.031 for our 50 quasars .
This result means that using power-law extrapolation we miss 5.3 % of flux on average and we show that there are cases where we would miss 16 % of flux .
These corrections are essential in the study of the intergalactic medium at high redshift in order to achieve precise measurements of physical properties , cosmological parameters , and the reionization epoch .