-
作者：DURBIN, J
摘要：A simple method of obtaining asymptotic expansions for the densities of sufficient estimators is described. It is an extension of the one developed by Barndorff-Nielsen and Cox (1979) for exponential families. A series expansion in powers of n-1 is derived of which the 1st term has an error of order n-1 which can effectively be reduced to n-3/2 by renormalization. The results obtained are similar to those given by Daniels''s (1954) saddlepoint method but the derivations are simpler. A brief tr...
-
作者：DURBIN, J
摘要：Approximations are found to the joint distributions of noncircular and circular partial serial correlation coefficients calculated from residuals from regression on Fourier series. Results for coefficients calculated from deviations from the true mean and from the sample mean are obtained as special cases. When the observations are independent the partial coefficients are approximately independently distributed in .beta.- distributions that are the same for all odd-order coefficients and the s...
-
作者：COX, DR
摘要：Non-Bayesian inference is considered for a scalar unknown parameter. When the minimal sufficient statistic is of dimension > 1 and no suitable exactly ancillary statistic is available, a statistic with a certain property of local ancillarity is calculated and the conditional distribution given that statistic is evaluated by Edgeworth expansion. The resulting distribution is used to calculate significance tests and confidence intervals which have desired probability levels to O(1/n), where n is...
-
作者：MUIRHEAD, RJ; WATERNAUX, CM
作者单位：Harvard University
摘要：An asymptotic theory for canonical correlation analysis is given for multivariate populations with finite-4th moments. Asymptomatic distributions of the sample canonical correlation coefficients and of statistics used for testing hypotheses about the population coefficients involve the 4th order cumulants of the parent population and are sensitive to departures from normality. These asymptotic distributions have surprisingly simple forms in the case of elliptical populations; here a modified t...
-
作者：NAGARSENKER, PB
摘要：The exact distribution of the likelihood ratio statistic for testing the equality of p 1-parameter exponential populations is obtained in a computational form. Tables of 5% and 1% significance points are given for p = 3(1)6. An asymptotic expansion in terms of a .beta.-distribution is also given. (Applicability to clinical trials in patients with acute disease and to accident data involving workers is discussed].
-
作者：RAY, SK; SAHAI, A
摘要：Two-parameter families of ratio-type and product-type estimators are proposed for a finite population mean based on simple random samples of observations on the variable of interest and a concomitant variable. The families contain estimators which have lower mean square error in practical situations than the usual ratio, product and sample mean estimators.
-
作者：COSTELLO, P; WOLFE, DA
摘要：A distribution-free treatments vs. control multiple comparisons procedure is proposed. It is based on the control sample median, and is appropriate when the control sample observations are easy and relatively inexpensive to collect, while the treatments data are costly and/or difficult to obtain. For this procedure the control sample has a fixed sample size and the treatments have random sample sizes determined by a truncated inverse binomial sequential sampling scheme. Properties of the proce...
-
作者：MUDHOLKAR, GS; TRIVEDI, MC
摘要：A normal approximation to the null distribution of the likelihood ratio statistic for testing the multivariate general linear hypothesis is developed. This approximation is compared with the well-known approximations due to Box (1949) and Rao (1948). The approximations due to Box and Rao deteriorate as the number of variables or, in the terminology of 1-way classification, the number of groups increases. Both approximations give poor results when the number of degrees of freedom for erros is s...
-
作者：BARNDORFFNIELSEN, O
摘要：Conditionality resolution, i.e., the construction of ancillary statistics and of expressions for the conditional distribution of the maximum likelihood estimate, is discussed for transformation models and exponential models, with the most emphasis on the latter. Such resolutions may be either exact or approximate. Several types of ancillaries are considered, a discussion is given of a general formula for the conditional distribution and several examples are presented.
-
作者：FORD, I; SILVEY, SD
摘要：Properties of a design constructed sequentially for a simple nonlinear problem are investigated both theoretically and by simulation. The design measure corresponding to the sequential design converges to an optimal measure, even when the latter has singular information matrix. The empirical study suggests that in repeated-sampling inference, the fact that the design is sequential can be effectively ignored.
