Abstract
Simpler methods have been introduced for statistical confidence test of radiation counter.In one of these methods the closer the unbiased estimate of the variance is to the unbiased estimate of the mean the counter is worthy of confidence. But these simpler methods cannot give the significance level, the power. of the test and the number of measurements necessary for obtaining the power.
To establish more precise method for testing, several types of methods are discussed in this paper. The “X2-goodness-of-fit test” is most familiar among them, but it has the demerit that the results of the test are fairly affected by the way of dividing the data into classes.The “modified X2-test” also has the same shortage as X2-test. Though the “ω2-test” is not affected by the way of dividing the data, it does not give the significance level.
It has been found through the study of comparison that the “likelihood ratio test” is the best for the “confidence test”. By using the last method, the critical region and the power curves of the confidence test are given in three diagrams. Lastly the counting results are plotted on the probability paper. This graph may visualize the goodness-of-fit, though it can't be given numerically. So it must become possible to estimate the goodness-of-fit from these plots on the probability paper through much experience.
