The concept of risk unbiasedness in statistical prediction

This paper extends the concept of risk unbiasedness for applying to statistical prediction and nonstandard inference problems, by formalizing the idea that a risk unbiased predictor should be at least as close to the “true” predictant as to any “wrong” predictant, on the average. A novel aspect of our approach is measuring closeness between a predicted value and the predictant by a regret function, derived suitably from the given loss function. The general concept is more relevant than mean unbiasedness, especially for asymmetric loss functions. For squared error loss, we present a method for deriving best (minimum risk) risk unbiased predictors when the regression function is linear in a function of the parameters. We derive a Rao–Blackwell type result for a class of loss functions that includes squared error and LINEX losses as special cases. For location-scale families, we prove that if a unique best risk unbiased predictor exists, then it is equivariant. The concepts and results are illustrated with several examples. One interesting finding is that in some problems a best unbiased predictor does not exist, but a best risk unbiased predictor can be obtained. Thus, risk unbiasedness can be a useful tool for selecting a predictor.     

Perfect aggregation of Bayesian analysis on compositional data

Sufficiency is a widely used concept for reducing the dimensionality of a data set. Collecting data for a sufficient statistic is generally much easier and less expensive than collecting all of the available data. When the posterior distributions of a quantity of interest given the aggregate and disaggregate data are identical, perfect aggregation is said to hold, and in this case the aggregate data is a sufficient statistic for the quantity of interest. In this paper, the conditions for perfect aggregation are shown to depend on the functional form of the prior distribution. When the quantity of interest is the sum of some parameters in a vector having either a generalized Dirichlet or a Liouville distribution for analyzing compositional data, necessary and sufficient conditions for perfect aggregation are also established.     

我国银行间债券市场现状与对策

介绍我国银行间债券市场的发展情况，指出大力发展银行间债券市场可以增强商业银行的竞争力，分别从提高资本充足率、增强贯彻货币政策能力、加强银行风险管理和完善做市商制度等方面详细阐述，并提出解决银行间债券现存问题的有效途径。     

On sequential estimation of the mean of a multidimensional gaussian process

For a multidimensional continuous time GAUssian process whose mean vector depends linearly of a multidimensional parameter, we consider a sequential estimation niodel. A suitable estimator process is constructed, and its distributions are analyzed in order to prove that it is progressively sufficient, After the reduction by sufficiency, an invariant structure is introduced and the optimal invariant terminal decision function is obtained.     

银行资产构成的特殊性与银行监管

为了论证银行监管的必要性,通过运用比较分析方法,对银行资产构成的特殊性进行剖析,得出银行是一个高负债进而高风险的行业客观结论,银行资产构成中92%的负债决定了银行监管是保护债权人利益的有效工具。     

Comparison of experiments for a class of positively dependent random variables

Using Blackwell's definition for comparison of experiments, it is shown that some sets of positively dependent random variables are less informative than similar sets of independent random variables. It is also shown that the information content of symmetric multivariate normal random vectors with a common known variance increases as the common correlation coefficient decreases. Some results which compare members of two-parameter exponential families are also included.     

THE REDUCTION OF RISKS FOR SET-VALUED ESTIMATORS

An extension of the Theorem of Rao-Blackwell will be given for the case of set-valued estimators, its images being nonempty, compact-convex subsets of a certain Banach space. Moreover sufficiency will be characterized in connection with set-valued estimators by related point-valued functions.     

On quantum statistical inference

Summary. Interest in problems of statistical inference connected to measurements of quantum systems has recently increased substantially, in step with dramatic new developments in experimental techniques for studying small quantum systems. Furthermore, developments in the theory of quantum measurements have brought the basic mathematical framework for the probability calculations much closer to that of classical probability theory. The present paper reviews this field and proposes and interrelates some new concepts for an extension of classical statistical inference to the quantum context.     

On a new type of sufficiency in undominated models

If a family of measures is dominated by a σ-finite measure, then the classical Fisher formulation of sufficiency and the pairwise sufficiency coincide. However, in undominated models, these definitions are essentially different. In particular, the Fisher formulation lacks the basic property: if a σ-algebra is sufficient, then also any larger σ-algebra ? is sufficient. We introduce a new definition of sufficiency, based on the concept of randomization, which has the property described above. We show that in the undominated case, our definition implies pairwise sufficiency and that the converse does not hold. If we assume that the underlying measurable space is a standard Borel space, then Fisher sufficiency implies our new formulation, and the converse implication does not hold.     

On principles and arguments to likelihood

Birnbaum (1962a) argued that the conditionality principle (C) and the sufficiency principle (S) implied the likelihood principle (L); he then argued (Birnbaum 1972) that C and a mathematical equivalence principle M implied L. Evans, Fraser, and Monette (1985a) gave reference details, and this paper gives proof that C alone implies L. The level of support by the profession for L is sharply less than that for S or even for C; thus the paradoxical nature of these results. In this regard, we elaborate on the Monette example (Fraser, Monette, and Ng 1984), which provides a strong case against L. We also examine closely the various proofs linking the principles and find that S and C can each be used operationally to suppress information otherwise deemed relevant. From another viewpoint this says that S and C can each be used in contexts that directly conflict with the original examples and motivations supporting them; the principles can thus be viewed as inappropriately used, or more strongly, as invalid. In either case, the result that C and S imply L or that C implies L can be regarded as noneffective in the context of discriminating applications. A resolution of the apparent anomalies can be obtained by allowing the statistical model to include ingredients additional to those usually present (particularly for subsequent use with conditionality), or alternatively by restricting the application of the principles to contexts where the conflicts would seem not to arise.