Geomechanical data are never sufficient in quantity or adequately precise and accurate for design purposes in mining and civil engineering. The objective of this paper is to show the variability of rock properties at the sampled point in the roadway’s roof, and then, how the statistical processing of the available geomechanical data can affect the results of numerical modelling of the roadway’s stability. Four cases were applied in the numerical analysis, using average values (the most common in geomechanical data analysis), average minus standard deviation, median, and average value minus statistical error. The study show that different approach to the same geomechanical data set can change the modelling results considerably. The case shows that average minus standard deviation is the most conservative and least risky. It gives the displacements and yielded elements zone in four times broader range comparing to the average values scenario, which is the least conservative option. The two other cases need to be studied further. The results obtained from them are placed between most favorable and most adverse values. Taking the average values corrected by statistical error for the numerical analysis seems to be the best solution. Moreover, the confidence level can be adjusted depending on the object importance and the assumed risk level.

对于采矿和土木工程中的设计目的，地质力学数据的数量永远不够或不够精确。本文的目的是显示巷道顶板采样点处岩石特性的可变性，然后说明可用地质力学数据的统计处理如何影响巷道稳定性数值模拟的结果。使用平均值（地质力学数据分析中最常见），平均值减去标准偏差，中位数和平均值减去统计误差，将四个案例应用于数值分析。研究表明，对相同的地质力学数据集采用不同的方法可以大大改变建模结果。案例表明，平均负标准差是最保守，风险最小的。与平均值方案相比，它提供的位移和屈服元素区域的范围是平均值的四倍，这是最不保守的选择。另外两种情况需要进一步研究。从它们获得的结果置于最有利和最不利值之间。采取统计误差校正的平均值进行数值分析似乎是最好的解决方案。此外，可以根据对象的重要性和假定的风险级别来调整置信度。采取统计误差校正的平均值进行数值分析似乎是最好的解决方案。此外，可以根据对象的重要性和假定的风险级别来调整置信度。采取统计误差校正的平均值进行数值分析似乎是最好的解决方案。此外，可以根据对象的重要性和假定的风险级别来调整置信度。