Abstract: Presently, the statistical methods are often used to delete abnormal Values in the environmental data manipulation. The majority of raw environmental data are those of the microcomponents obtained in the environment monitoring. According to the present knowledge, the frequency distribution of a microcompo-nent concentration in the environment is most likely to be log-normal or extreme positive skew distribution instead of a normal distribution. However, the statistical methods usually used to delete abnormal values are designed for the samples with normal distribution. Consequently, the use of these statistical methods often tead to an accuracy reduction of the data because of the unreasonable deletion of the high values in the sample. Using some examples for the microcomponeuts in the soil, this article revealed how the usually used GRUBBS method had brought a distortron to those examples. This artical also showed that the test of normality could be used to diminish or to avoid such distortion.