Abstract: In this paper, we consider the Bayesian posterior distribution as the solution to a minimization rule, first observed by Zellner (1988). The expression to be minimized is a mixture of two pieces, one piece involving the prior distribution, which is minimized by the prior, and the other piece involves the data, which is minimized by the measure putting all the mass on the maximum likelihood estimator. From this perspective of the posterior distribution, Bayesian model selection and the search for an objective prior distribution, can be viewed in a way which is different from usual Bayesian approaches.

Abstract: In this paper we present asymptotic estimates of level crossing probabilities from a Bayesian point of view, based on large deviations. For the Bayesian analysis we choose a finite mixture of conjugate prior distributions to model the uncertainty on the unknown parameters of the two classes of stochastic processes considered: the Brownian motion and the compound Poisson process with upward jumps and negative drift. The estimates of level crossing probabilities are derived as a consequence of large deviation principles for posterior distributions.

Abstract: Minimum bias (all bias) designs for the linear model were proposed by Box and Draper. In this article we extend their results to generalized linear models. We show that, in the canonical case, the minimum bias design density is: (a) proportional to the weight function reflecting the importance of each design point, and (b) inversely proportional to the observation variance at each point. We also derive minimum bias design densities in non-canonical cases. Implications for binary, Poisson and exponential data are considered in the examples. From these examples we observe that when the experimenter is mainly interested in the mean of the Binomial, Poisson or exponential distribution, rather than the canonical parameter, and if the weight function is chosen to be inversely proportional to the variance of the maximum likelihood estimator of the mean, minimum bias designs are uniform. These uniform designs automatically minimize the bias/standard error ratio and the mean square error/variance ratio.

Abstract: Consider a sequence of independent and identically distributed random variables with the underlying distribution in the domain of attraction of a stable distribution with an exponent in (0;2]. Deflne the trimmed sums as the partial sums excluding r largest observations in magnitude, where r is a flxed integer. This paper proves that Chover’s law of the iterated logarithm holds for the trimmed sums.

Abstract: The problem of estimating the parameters of multivariate linear models in the context of an arbitrary pattern of missing data is addressed in the present paper. While this problem is frequently handled by EM strategies, we propose a Gauss-Markov approach based on an initial linearization of the covariance of the model. A complete class of quadratic estimators is first exhibited in order to derive locally Minimum Variance Quadratic Unbiased Estimators (MIVQUE) of the variance parameters. Apart from the interest in locally MIVQUE itself, this approach gives more insight into maximum likelihood estimation. Indeed, an iterated version of MIVQUE is proposed as an alternative to EM to calculate the maximum likelihood estimators. Finally, MIVQUE and maximum likelihood estimation are compared by simulations.