Abstract: Laser induced breakdown spectroscopy (LIBS) is an atom analysis technique which can be used for studying the element composition of multitudinous specimen matrices, including gas, liquid or solid.LIBS has demonstrated significant potential for the gemstone identification, with the capabilities of simultaneous multiple elements detection and rapid analysis.However, for the unstable ablation, matrix effect and self-absorption, there has been long controversy about its quantitative analytical capability.Univariate and multivariate analysis are the main means for LIBS to analyze and test element composition of multitudinous specimen matrices.Multivariate analytical techniques have great potential for analyzing the complicated LIBS spectra.In this paper, univariate and multivariate statistical analysis techniques were coupled with LIBS to identify different types of jadeite samples with various elemental composition concentrations.The authors employed the univariate and the partial least square (PLS) techniques to analyze the LIBS spectra of jadeite samples and built calibration models predicting Ca and Al concentrations.The results indicated that PLS could significantly improve the analytical results in comparison with the univariate technique.The normalized root mean square error (RMSE) and R² of the univariate models were 1.14 and 0.93 in calibration and 1.43 and 0.92 in prediction for Ca, while 3.04 and 0.63 in calibration and 3.59 and 0.58 in prediction for Al.For the PLS models using the spectral range 197.33 to 762.8 nm, the RMSE and R² are 0.81 and 0.97 for Ca, 1.08 and 0.95 for Al in calibration respectively, as well as 1.36 and 0.92 for Ca and 1.79 and 0.89 for Al in prediction respectively.The results showed that PLS could enhance the quantitative analytical capability of LIBS for jadeite analysis.