Abstract: In order to obtain more reliable results in differential analysis between test value μ and sensory evaluation data of tobacco and tobacco products, the outliers of sensory evaluation data were rejected, the significance of difference between μ and sensory evaluation data was judged based on mean value confidence interval, and the minimum difference was determined. The results showed that:1) Comparing with mean value method, this method effectively improved the reliability of analysis results. 2) The change trend and extent of the sensory evaluation data of tobacco and tobacco products after a certain treatment could be scientifically determined with this method, which effectively decreased the influences of outliers on analysis results. This method provided a reference for the statistical analysis of data in the evaluating flavor' s function, effect of processing on tobacco sensory quality and result of cigarette improvement, etc.