Bayesian model selection using test statistics

Summary.  Existing Bayesian model selection procedures require the specification of prior distributions on the parameters appearing in every model in the selection set. In practice, this requirement limits the application of Bayesian model selection methodology. To overcome this limitation, we propose a new approach towards Bayesian model selection that uses classical test statistics to compute Bayes factors between possible models. In several test cases, our approach produces results that are similar to previously proposed Bayesian model selection and model averaging techniques in which prior distributions were carefully chosen. In addition to eliminating the requirement to specify complicated prior distributions, this method offers important computational and algorithmic advantages over existing simulation-based methods. Because it is easy to evaluate the operating characteristics of this procedure for a given sample size and specified number of covariates, our method facilitates the selection of hyperparameter values through prior-predictive simulation.     

Distribution-free tests in randomized blocks

Classes of distribution-free tests are proposed for testing homogeneity against order restricted as well as unrestricted alternatives in randomized block designs with multiple observations per cell. Allowing for different interblock scoring schemes, these tests are constructed based on the method of within block rankings. Asymptotic distributions (cell sizes tending to infinity) of these tests are derived under the assumption of homogeneity. The Pitman asymptotic relative efficiencies relative to the least squares statistics are studied. It is shown that when blocks are governed by different distributions, adaptive choice of scores within each block results in asymptotically more efficient tests as compared with methods that ignore such information. Monte Carlo simulations of selected designs indicate that the method of within block rankings is more power robust with respect to differing block distributions.     

Equivalent inferences using transformations

Justice (1977) has presented a Leviiison-typc- solution for two dimensional Wiener filtering problem. Since the solution is based on the Szego polynomials, it is necessary to calculate the bivariate Szego polynomials to obtain it In this paper, an alternative solution of the problem is proposed, which is based on the block Cholesky decomposition of the inverse of a Hermi-tian block Toeplitz matrix. Since the block Choiesky decomposition can be accomplished through the Whittle algorithm, the new solution is easy to implement into a computer program.     

Confidence intervals for the ratio of two variances

In this paper we consider confidence intervals for the ratio of two population variances. We propose a confidence interval for the ratio of two variances based on the t-statistic by deriving its Edgeworth expansion and considering Hall's and Johnson's transformations. Then, we consider the coverage accuracy of suggested intervals and intervals based on the F-statistic for some distributions.     

The limiting distribution of the F-statistic from nonnormal universes

We consider a linear regression model with an unbalanced 1-fold nested error structure, where group effect and error are from nonnormal universes. The limiting distribution of the F-statistic in this model is derived, as the sample size is large and group sizes take values from a finite set of distinct integers. The result is used to approximate the F-distribution quantile and to test the significance of the random effect variance component. Results are also applicable to the F-statistic in the one-way random-effects model. The effects of departure from normality on the F-statistic distribution are given.     

Test procedures for the analysis of experimental designs with correlated nonnormal data

Using Monte Carlo methods, an examination is made of two statistical methods used for hypothesis testing in a general factorial model with a known correlation structure General correlation structures are given in Smith and Lewis (1980) and Pavur and Lewis (1982) which allow the usual F statistic to be corrected by a constant. The corrected F statistic would be the usual F statistic multiplied by a correction constant. A comparison is made between this corrected f statistic and the rank transform F statistic presented by Conover and Iman (1976). When the usual F statistic and the rank transform statistic are corrected for correlationt this simulation study shows that these statistical tests behave well under a variety of situations when not all f the usual assumptions of an ANOVA are satisfied.