Simulations from Laboratory Sourceless Object Counting System (LabSOCS) software were used to determine self-attenuation correction factor, which is defined as the efficiency ratio of the sample with the absorbing medium to that of the sample without absorbing medium. The semi-empirical self-attenuation correction formula F(μ) used to correct self-attenuation of a sample was applied. A comparison of the two methods reveals that formula of sample with φ75 mm×25 mm and φ75 mm×10 mm can be, respectively, used in the self-attenuation correction for μ in the ranges of 0 to 0.5 cm−1 and 0.5 cm−1 to 2.0 cm−1, indicating that the semi-empirical formula will not be used when μ has exceeded the interval. The semi-empirical formula value is consistent with the experimental value, within 7.9% accuracy. Therefore, this method is correct and effective. Both of our two methods can accurately produce a relative self-attenuation correction factor when the composition of the sample is known. The self-attenuation correction of a sample with unknown composition can only be carried out using a semi-empirical formula method.