On Limiting Distribution of Quasi-Posteriors under Partial Identification

Wenxin Jiang

DOI: https://doi.org/10.48550/arXiv.1610.01747

2016-10-06

Abstract:We establish the limiting distribution (in total variation) of the quasi posteriors based on moment conditions, which only partially identify the parameters of interest. Some examples are discussed.

Statistics Theory

What problem does this paper attempt to address?

The Variance of Partial Sums of Strong Near-Epoch Dependent Variables

Jin Qiu,Zhengyan Lin

DOI: https://doi.org/10.1016/j.spl.2006.04.031

IF: 0.718

2006-01-01

Statistics & Probability Letters

Abstract:In this paper, we discuss the lower bound for the variance of partial sums of strong near-epoch dependent variables under quite weak conditions. By suitably applying this result, we derive the limit behavior of the variance of the partial sums, which is especially useful in studying the large sample properties for econometric time series models with strong near-epoch dependent innovations.
Moments and Limiting Distribution of a Portfolio of Whole Life Annuity Policies

He Wen-jiong,Zhang Yi

DOI: https://doi.org/10.1007/bf02839489

2002-01-01

Journal of Zhejiang University SCIENCE A

Abstract:A dual random model of a portfolio of variable amount whole life annuity is set with the m th moment of the present value of benefits, and the respective expressions of the moments under the assumption that the force of interest accumulation function is Wiener process or Ornstein-Uhlenbeck process. Furthermore, the limiting distribution of average cost of this portfolio is discussed with the expression of the limiting distribution under the assumption that the force of interest accumulation is an independent increment process.
On the Limiting Behaviors of Increments of Sums of Random-Variables Without Moment Conditions

ZY LIN,QM SHAO

1993-01-01

Chinese Annals of Mathematics Series B

Abstract:The a.s. limiting behaviors of big increments of partial sums of a sequence of random variables are obtained without moment conditions.The theorems sharpen the results of Lin[5].
On Increments of Sums of Random-Variables Without Moment Hypotheses

ZY LIN

DOI: https://doi.org/10.1360/ya1990-33-9-1047

1990-01-01

Abstract:> The a. s. limiting behavior of increments of partial sums for a sequence of random variables has been discussed by many authors. A series of fine and elegant results have been obtained. However, up to now, the results of this kind have all required the existence of moments. In this paper we try to drop this condition. The existing results with moment conditions become special cases of our theorems.
Almost Sure Max-Limits for Nonstationary Gaussian Sequence

SQ Chen,ZY Lin

DOI: https://doi.org/10.1016/j.spl.2005.12.018

IF: 0.718

2006-01-01

Statistics & Probability Letters

Abstract:We obtain some almost sure limit theorems for a standardized nonstationary Gaussian sequence under some mild conditions.
LIMIT BEHAVIORS FOR THE INCREMENTS OF A D-Dimensional MULTI-PARAMETER GAUSSIAN PROCESS

YK Choi,ZY Lin,HS Sung,KS Hwang,HJ Moon

DOI: https://doi.org/10.4134/jkms.2005.42.6.1265

2005-01-01

Journal of the Korean Mathematical Society

Abstract:In this paper, we establish limit theorems containing both the moduli of continuity and the large incremental results for finite dimensional Gaussian processes with N parameters, via estimating upper bounds of large deviation probabilities on suprema of the Gaussian processes.
Limit Theorems for Dependent Bernoulli Variables with Statistical Inference

Yang Zhang,Lixin Zhang

DOI: https://doi.org/10.1080/03610926.2015.1021015

2017-01-01

Abstract:We consider a class of dependent Bernoulli variables where the conditional success probability is a linear combination of the last few trials and the original success probability. We obtain its limit theorems including the strong law of large numbers, weak invariance principle, and law of the iterated logarithm. We also derive some statistical inference results which make the model applicable. Simulation results are exhibited as well to show that with small sample size the convergence rate is satisfying and the proposed estimators behave well.
SELF-NORMALIZED CENTRAL LIMIT THEOREM AND ESTIMATION OF VARIANCE OF PARTIAL SUMS FOR NEGATIVE DEPENDENT RANDOM VARIABLES

Zhang Lixin,Shi Strongway

DOI: https://doi.org/10.1007/s11766-002-0012-z

2002-01-01

Abstract:Let { X n , n≥1} be a stationary LNQD or NA sequence satisfying EX 1 = μ, EX 1 2 <∞ and (Var S n )/ n →σ 2 as n →∞. In this paper a class of self-normalized central limit theorems and estimators of Var S n are studied. The weak and strong consistency of the estimators of Var S n are presented.
On the almost sure invariance principle for dependent Bernoulli random variables

yang zhang,lixin zhang

DOI: https://doi.org/10.1016/j.spl.2015.09.008

IF: 0.718

2015-01-01

Statistics & Probability Letters

Abstract:We consider a sequence of dependent Bernoulli variables where the success probability of the trial conditional on the past history is a linear function of the mean number of successes achieved to that point. An almost sure invariance principle is established for the partial sum and we also generalize the model to a multi-dimensional case, extending the results of Heyde (2004), James et al. (2008) and Wu et al. (2012).
Vanishing of quasi-invariant generalized functions

dihua jiang,binyong sun,chenbo zhu

2012-01-01

Abstract: Determination of quasi-invariant generalized functions is important for a variety of problems in representation theory, notably character theory and restriction problems. In this note, we review some new and easy-to-use techniques to show vanishing of quasi-invariant generalized functions, developed in the recent work of the authors (Uniqueness of Ginzburg-Rallis models: the Archimedean case, Trans. Amer. Math. Soc. 363, (2011), 2763-2802). The first two techniques involve geometric notions attached to submanifolds, which we call metrical properness and unipotent $\chi$-incompatibility. The third one is analytic in nature, and it arises from the first occurrence phenomenon in Howe correspondence. We also highlight how these techniques quickly lead to two well-known uniqueness results, on trilinear forms and Whittaker models.
Inference in Semiparametric Conditional Moment Models with Partial Identification

Shengjie Hong

DOI: https://doi.org/10.1016/j.jeconom.2016.09.014

IF: 3.363

2016-01-01

Journal of Econometrics

Abstract:This paper develops inference methods for conditional moment models in which the unknown parameter is possibly partially identified and may contain infinite-dimensional components. For a conjectured restriction on the parameter, we consider testing the hypothesis that the restriction is satisfied by at least one element of the identified set. We propose using the sieve minimum of a Kolmogorov–Smirnov type statistic as the test statistic, derive its asymptotic distribution, and provide consistent bootstrap critical values. In this way a broad family of restrictions can be consistently tested, making the proposed procedure applicable to testing the model specification and constructing confidence set for any given component or some feature of the parameter. Our methods are robust to partial identification, and allow for the moment functions to be nonsmooth. As an illustration, we apply the proposed inference methods to study the quantile instrumental variable Engel curves for gasoline in Brazil. A Monte Carlo study demonstrates finite sample performance.
The limit theorems on extreme order statistics and partial sums of i.i.d. random variables

Gaoyu Li,Zhongquan Tan

2023-12-15

Abstract:This paper proves several weak limit theorems for the joint version of extreme order statistics and partial sums of independently and identically distributed random variables. The results are also extended to almost sure limit version.

Probability
Quasi-negative binomial distribution: Properties and applications

Shubiao Li,Fang Yang,Felix Famoye,Carl Lee,Dennis Black

DOI: https://doi.org/10.1016/j.csda.2011.02.003

2011-07-01

Abstract:In this paper, a quasi-negative binomial distribution (QNBD) derived from the class of generalized Lagrangian probability distributions is studied. The negative binomial distribution is a special case of QNBD. Some properties of QNBD, including the upper tail behavior and limiting distributions, are investigated. It is shown that the moments do not exist in some situations and the limiting distribution of QNBD is the generalized Poisson distribution under certain conditions. A zero-inflated QNBD is also defined. Applications of QNBD and zero-inflated QNBD in various fields are presented and compared with some other existing distributions including Poisson, generalized Poisson and negative binomial distributions as well as their zero-inflated versions. In general, the QNBD or its zero-inflated version performs better than the other models based on the chi-square statistic and the Akaike Information Criterion, especially for the cases where the data are highly skewed, have heavy tails or excessive numbers of zeros.

statistics & probability,computer science, interdisciplinary applications
Posterior distribution existence and error control in Banach spaces in the Bayesian approach to UQ in inverse problems

J. Andrés Christen,Marcos A. Capistrán,M. Luisa Daza-Torres,Hugo Flores-Argüedas,J. Cricelio Montesinos-López

DOI: https://doi.org/10.48550/arXiv.1712.03299

2018-10-13

Abstract:We generalize the results of \cite{Capistran2016} on expected Bayes factors (BF) to control the numerical error in the posterior distribution to an infinite dimensional setting when considering Banach functional spaces and now in a prior setting. The main result is a bound on the absolute global error to be tolerated by the Forward Map numerical solver, to keep the BF of the numerical vs. the theoretical model near to 1, now in this more general setting, possibly including a truncated, finite dimensional approximate prior measure. In so doing we found a far more general setting to define and prove existence of the infinite dimensional posterior distribution than that depicted in, for example, \cite{Stuart2010}. Discretization consistency and rates of convergence are also investigated in this general setting for the Bayesian inverse problem.

Statistics Theory
On the Posterior Distribution in Denoising: Application to Uncertainty Quantification

Hila Manor,Tomer Michaeli

2024-02-19

Abstract:Denoisers play a central role in many applications, from noise suppression in low-grade imaging sensors, to empowering score-based generative models. The latter category of methods makes use of Tweedie's formula, which links the posterior mean in Gaussian denoising (\ie the minimum MSE denoiser) with the score of the data distribution. Here, we derive a fundamental relation between the higher-order central moments of the posterior distribution, and the higher-order derivatives of the posterior mean. We harness this result for uncertainty quantification of pre-trained denoisers. Particularly, we show how to efficiently compute the principal components of the posterior distribution for any desired region of an image, as well as to approximate the full marginal distribution along those (or any other) one-dimensional directions. Our method is fast and memory-efficient, as it does not explicitly compute or store the high-order moment tensors and it requires no training or fine tuning of the denoiser. Code and examples are available on the project webpage in <a class="link-external link-https" href="https://hilamanor.github.io/GaussianDenoisingPosterior/" rel="external noopener nofollow">this https URL</a> .

Computer Vision and Pattern Recognition,Machine Learning
Sequential Gibbs Posteriors with Applications to Principal Component Analysis

Steven Winter, Omar Melikechi, David B. Dunson

2023-10-20

Abstract:Gibbs posteriors are proportional to a prior distribution multiplied by an exponentiated loss function, with a key tuning parameter weighting information in the loss relative to the prior and providing a control of posterior uncertainty. Gibbs posteriors provide a principled framework for likelihood-free Bayesian inference, but in many situations, including a single tuning parameter inevitably leads to poor uncertainty quantification. In particular, regardless of the value of the parameter, credible regions have far from the nominal frequentist coverage even in large samples. We propose a sequential extension to Gibbs posteriors to address this problem. We prove the proposed sequential posterior exhibits concentration and a Bernstein-von Mises theorem, which holds under easy to verify conditions in Euclidean space and on manifolds. As a byproduct, we obtain the first Bernstein-von Mises theorem for traditional likelihood-based Bayesian posteriors on manifolds. All methods are illustrated with an application to principal component analysis.

Machine Learning,Methodology
Bayesian Quantile Estimation and Regression with Martingale Posteriors

Edwin Fong,Andrew Yiu

2024-06-05

Abstract:Quantile estimation and regression within the Bayesian framework is challenging as the choice of likelihood and prior is not obvious. In this paper, we introduce a novel Bayesian nonparametric method for quantile estimation and regression based on the recently introduced martingale posterior (MP) framework. The core idea of the MP is that posterior sampling is equivalent to predictive imputation, which allows us to break free of the stringent likelihood-prior specification. We demonstrate that a recursive estimate of a smooth quantile function, subject to a martingale condition, is entirely sufficient for full nonparametric Bayesian inference. We term the resulting posterior distribution as the quantile martingale posterior (QMP), which arises from an implicit generative predictive distribution. Associated with the QMP is an expedient, MCMC-free and parallelizable posterior computation scheme, which can be further accelerated with an asymptotic approximation based on a Gaussian process. Furthermore, the well-known issue of monotonicity in quantile estimation is naturally alleviated through increasing rearrangement due to the connections to the Bayesian bootstrap. Finally, the QMP has a particularly tractable form that allows for comprehensive theoretical study, which forms a main focus of the work. We demonstrate the ease of posterior computation in simulations and real data experiments.

Methodology,Statistics Theory
Point-wise characterizations of limits of planar Sobolev homeomorphisms and their quasi-monotonicity

Daniel Campbell

2024-01-22

Abstract:We present three novel classifications of the weak sequential (and strong) limits in $W^{1,p}$ of planar diffeomorphisms. We introduce a concept called the QM condition which is a kind of separation property for pre-images of closed connected sets and show that $u$ satisfies this property exactly when it is the limit of Sobolev homeomorphisms. Further, we prove that $u\in W^{1,p}_{\operatorname{id}}((-1,1)^2,\mathbb{R}^2)$ is the limit of a sequence of homeomorphisms exactly when there are classically monotone mappings $g_{\delta}:[-1,1]^2\to \mathbb{R}^2$ and very small open sets $U_{\delta}$ such that $g_{\delta} = u$ on $[-1,1]^2 \setminus U_{\delta}$. Also, we introduce the so-called three curve condition, which is in some sense reminiscent of the NCL condition of \cite{CPR} but for $u^{-1}$ instead of for $u$, and prove that a map is the $W^{1,p}$ limit of planar Sobolev homeomorphisms exactly when it satisfies this property. This improves on results in \cite{DPP} answering the question from \cite{IO2}.

Analysis of PDEs
Are minimizers of the Onsager-Machlup functional strong posterior modes?

Remo Kretschmann

DOI: https://doi.org/10.1137/23M1546579

2023-07-03

Abstract:In this work we connect two notions: That of the nonparametric mode of a probability measure, defined by asymptotic small ball probabilities, and that of the Onsager-Machlup functional, a generalized density also defined via asymptotic small ball probabilities. We show that in a separable Hilbert space setting and under mild conditions on the likelihood, modes of a Bayesian posterior distribution based upon a Gaussian prior exist and agree with the minimizers of its Onsager-Machlup functional and thus also with weak posterior modes. We apply this result to inverse problems and derive conditions on the forward mapping under which this variational characterization of posterior modes holds. Our results show rigorously that in the limit case of infinite-dimensional data corrupted by additive Gaussian or Laplacian noise, nonparametric maximum a posteriori estimation is equivalent to Tikhonov-Phillips regularization. In comparison with the work of Dashti, Law, Stuart, and Voss (2013), the assumptions on the likelihood are relaxed so that they cover in particular the important case of white Gaussian process noise. We illustrate our results by applying them to a severely ill-posed linear problem with Laplacian noise, where we express the maximum a posteriori estimator analytically and study its rate of convergence in the small noise limit.

Statistics Theory,Optimization and Control

On Limiting Distribution of Quasi-Posteriors under Partial Identification

The Variance of Partial Sums of Strong Near-Epoch Dependent Variables

Moments and Limiting Distribution of a Portfolio of Whole Life Annuity Policies

On the Limiting Behaviors of Increments of Sums of Random-Variables Without Moment Conditions

On Increments of Sums of Random-Variables Without Moment Hypotheses

Almost Sure Max-Limits for Nonstationary Gaussian Sequence

LIMIT BEHAVIORS FOR THE INCREMENTS OF A D-Dimensional MULTI-PARAMETER GAUSSIAN PROCESS

Limit Theorems for Dependent Bernoulli Variables with Statistical Inference

SELF-NORMALIZED CENTRAL LIMIT THEOREM AND ESTIMATION OF VARIANCE OF PARTIAL SUMS FOR NEGATIVE DEPENDENT RANDOM VARIABLES

On the almost sure invariance principle for dependent Bernoulli random variables

Vanishing of quasi-invariant generalized functions

Inference in Semiparametric Conditional Moment Models with Partial Identification

The limit theorems on extreme order statistics and partial sums of i.i.d. random variables

Quasi-negative binomial distribution: Properties and applications

Posterior distribution existence and error control in Banach spaces in the Bayesian approach to UQ in inverse problems

On the Posterior Distribution in Denoising: Application to Uncertainty Quantification

Sequential Gibbs Posteriors with Applications to Principal Component Analysis

Bayesian Quantile Estimation and Regression with Martingale Posteriors

Point-wise characterizations of limits of planar Sobolev homeomorphisms and their quasi-monotonicity

Are minimizers of the Onsager-Machlup functional strong posterior modes?