Along with the inherent scattering in soil physical and mechanical parameters of laboratory testing, it is difficult for students to sustain their conviction in their own tests. It is necessary to explore how to train students' uncertainty thinking through soil mechanics laboratory testing. To improving teaching effect, statistical characteristics of soil parameters are expounded by the drawing and scientific computing functions of statistical programming language R. Multiple parallel experiments on various parameters, including particle size analysis, moisture content, liquid and plastic limits, permeability coefficient, compression coefficient and shear strength, are used to demonstrate how uncertainties can be quantified. Test results are presented using probability density distribution curves, non-parametric kernel density estimates, histograms, and box plots. Such teaching practice can enhance undergraduates' comprehensive judgment and inductive ability in handling inconsistent data.