To get accurate background value and anomaly threshold, it is necessary to iterate abnormally distributed geochemical data for deleting outliers in geochemistry. To analyze deeply the complex iteration procedure, actual data and "outliers check" the authors used iteration procedure to detect the effect of different parameters and reliability of logarithm transformation. Some conclusions have been reached: According to the goal and 3σrule of normal distribution, in x_i±kS_i,k=3 is suitable. The higher anomaly threshold of logarithm transformation is not correct and the true value of logarithm standard deviation is not in accord with the standard deviation of the true value, so the the threshold is equal to the sum of geometric mean and standard deviation of the true value.