-
作者:Newman, D
摘要:The object of the paper is to consider the ratio q[long dash]w/s, where w is the range of a sample of n individuals from a normal population, and s is an independent estimate of the population standard deviation with f/ degrees of freedom. The expectation of q is first considered and a short table given, by use of which, in conjunction with existing tables, this expectation can be calculated. The methods of calculating the upper 5% and 1% levels of the distribution of q (exact for n=2, 3, appr...
-
作者:Welch, BL
摘要:A method of approximating the probability integral of the distribution of a maximum likelihood estimate is given.
-
作者:Pitman, EJG
摘要:Suppose that X and Y are normally correlated variables and that we have a sample of n pairs of observed values (x r, y r) and wish to determine whether the sample values indicate that the variances of X and Y are unequal. A simple and exact significance test is developed, which may be used when the exact value of the correlation coefficient of X and Y is unknown. It is also shown how fiducial limits for the ratio of the variances of X and Y may be determined. A similar method gives very easily...
-
作者:Pitman, EJG
摘要:Tests of certain hypotheses concerning location and scale parameters of chance variables are developed. These tests are unbiased and are applicable to chance variables with any continuous distributions. An application to the comparison of variances of normal variables yields Bartlett''s test, which is thus shown to be unbiased. This test is discussed in detail. The Neyman-Pearson variances test is shown to be biased except when the samples are all of the same size.
-
作者:Welch, BL
摘要:The problem of discriminating between two specified populations is discussed.