Comparison of least-squares vs. maximum likelihood estimation for standard spectrum technique of beta-gamma coincidence spectrum analysis

作者：Lowrey Justin D^*; Biegalski Steven R F

来源：Nuclear Instruments and Methods in Physics Research Section B: Beam Interactions with Materials and Atoms , 2012, 270: 116-119.

DOI：10.1016/j.nimb.2011.09.005

• Deconvolution
• Maximum likelihood
• Low-counting statistics

The spectrum deconvolution analysis tool (SDAT) software code was written and tested at The University of Texas at Austin utilizing the standard spectrum technique to determine activity levels of Xe-131m, Xe-133m, Xe-133, and Xe-135 in beta-gamma coincidence spectra. SDAT was originally written to utilize the method of least-squares to calculate the activity of each radionuclide component in the spectrum. Recently, maximum likelihood estimation was also incorporated into the SDAT tool. This is a robust statistical technique to determine the parameters that maximize the Poisson distribution likelihood function of the sample data. In this case it is used to parameterize the activity level of each of the radioxenon components in the spectra. A new test dataset was constructed utilizing Xe-131m placed on a Xe-133 background to compare the robustness of the least-squares and maximum likelihood estimation methods for low counting statistics data. The Xe-131m spectra were collected independently from the Xe-133 spectra and added to generate the spectra in the test dataset. The true independent counts of Xe-131m and Xe-133 are known, as they were calculated before the spectra were added together. Spectra with both high and low counting statistics are analyzed. Studies are also performed by analyzing only the 30 key X-ray region of the beta-gamma coincidence spectra. Results show that maximum likelihood estimation slightly outperforms least-squares for low counting statistics data.

• 出版日期2012-1-1