Using a high luminosity ($L_{\rm bol} \sim 10^{47.5} - 10^{48.3}$ erg s$^{-1}$), high redshift ($3.2 < z < 3.8$) quasar sample of 19 quasars with optical and near-infrared spectroscopy, we investigate the reliability of the CIV-based black hole mass estimates ($M_{\rm BH}$). The median logarithm of the CIV- and H$\beta$-based $M_{\rm BH}$ ratios is 0.110 dex with the scatter of 0.647 dex. The CIV-to-H$\beta$ BH mass differences are significantly correlated with the CIV FWHMs, blueshifts and asymmetries. Corrections of the CIV FWHM using the blueshift and asymmetry reduce the scatter of the mass differences by $\sim$ 0.04-0.2 dex. Quasars in our sample accrete at the Eddington ratio $R_{\rm EDD}>0.3$ and cover a considerable range of blueshifts, with 18/19 of the quasars showing CIV blueshifts (with the median value of 1126 km s$^{-1}$) and 14/19 of the quasars showing CIV blueshifts larger than 500 km s$^{-1}$. It suggests that not all quasars with high Eddington ratios show large blueshifts. The Baldwin effect between the CIV rest-frame equivalent width (REW) and the continuum luminosity at 1350 $\overset{\lower.5em\circ}{\mathrm{A}}$ is not seen, likely due to the limited luminosity range of our sample. We find a lack of flux in the red wing of the composite spectrum with larger CIV blueshift, and detect a higher ratio of [OIII] quasars with REW$_{\rm [OIII]}>5$ $\overset{\lower.5em\circ}{\mathrm{A}}$ in the subsample with lower CIV blueshift. It is more likely that they are caused by the combination of the Eddington ratio and the orientation effect.
Comment: 17 pages, 12 figures with additional 6 pages and 2 figures in the appendix. Accepted for publication in ApJ