Assessment of Random Error in Phantom Dosimetry with the Use of Error Simulation in Statistical Software

OBJECTIVE To investigate if software simulation is practical for quantifying random error (RE) in phantom dosimetry. MATERIALS AND METHODS We applied software error simulation to an existing dosimetry study. The specifications and the measurement values of this study were brought into the software (R version 3.0.2) together with the algorithm of the calculation of the effective dose (E). Four sources of RE were specified: (1) the calibration factor; (2) the background radiation correction; (3) the read-out process of the dosimeters; and (4) the fluctuation of the X-ray generator. RESULTS The amount of RE introduced by these sources was calculated on the basis of the experimental values and the mathematical rules of error propagation. The software repeated the calculations of E multiple times (n = 10,000) while attributing the applicable RE to the experimental values. A distribution of E emerged as a confidence interval around an expected value. CONCLUSIONS Credible confidence intervals around E in phantom dose studies can be calculated by using software modelling of the experiment. With credible confidence intervals, the statistical significance of differences between protocols can be substantiated or rejected. This modelling software can also be used for a power analysis when planning phantom dose experiments.