Using the method of multivariate statistical analysis, we have conducted the partial least squares regression analysis(PLSRA) of 8 kinds of alkaloids of 124 flue-cured tobacco samples in 2010 and 2011 in Guizhou. The result shows that nornicotine, nicotyrine, anatabine and cotinine have the greatest influence on nicotine content. By partial least squares regression equation to predict the content of nicotine. We have found that the relative standard deviation between predicted value of 119 samples of the training set and their measured value is less than 17.04% and that in 109 of them is less than 10.00%. At last, 5 samples of the detecting set have been applied to validate the partial least squares regression equation. The result shows that the relative standard deviation between predicted value and measured value is less than 6.72% and a good fitting has been obtained.