Optimal online detection of parameter changes in two linear models

Shiryaev has obtained the optimal sequential rule for detecting the instant of a distributional change in an independent sequence using the theory of optimal stopping of Markov processes. This paper considers the problem of sequential detection of certain parameter changes in two dependent sequences: an autoregressive process, and a regression model with serially correlated error terms. It is shown that the rule that is optimal in the sense of minimizing the expected positive delay is the one which declares a change to have occured as soon as the posterior probability of a change crosses a threshold. This rule also permits control of the probability of a false-declaration of change, just as in the independent sequence case.     

On some growth models with a small parameter

Summary We consider the behavior of the asymptotic speed of growth and the asymptotic shape in some growth models, when a certain parameter becomes small. The basic example treated is the variant of Richardson's growth model on ^d in which each site which is not yet occupied becomes occupied at rate 1 if it has at least two occupied neighbors, at rate 1 if it has exactly 1 occupied neighbor and, of course, at rate 0 if it has no occupied neighbor. Occupied sites remain occupied forever. Starting from a single occupied site, this model has asymptotic speeds of growth in each direction (as time goes to infinity) and these speeds determine an asymptotic shape in the usual sense. It is proven that as tends to 0, the asymptotic speeds scale as ^1/d and the asymptotic shape, when renormalized by dividing it by ^1/d , converges to a cube. Other similar models which are partially oriented are also studied.The work of R.H.S. was supported by the N.S.F. through grant DMS 91-00725. In addition, both authors were supported by the Newton Institute in Cambridge. The authors thank the Newton Institute for its support and hospitality     

Estimation of parameters in ARUMA models

Let X(n) be a time series satisfying the following ARUMA(p, d, q) models:U (B) A (B)X (n)=C (B) W (n)where U(B)=1+u(1)B+…+u(d) B~d is a polynomial with all roots on the unit circle, A(B)=1+a(1)B+…+a(p)Bp is a polynomial with all roots outside the unit circle, C(B)=1+c(1) B+…+c(q)Bq is a polynomial which is relatively prime with the polynomial U(B)A(B), B is thebackshift operator such that BX(n)=X(n-1), and (W (n), F(n), n≥1) is a sequence of martingaledifferences satisfying the following conditions:lim E (W (n)~2|F(n-1))=σ~2 a.s.n→∞sup E |W(n)|γ<∞ for some γ>2.n≥1The purpose of this paper is to provide consistent estimates of the parameters p, d, q, u(j) (j=1,2,…,d), and a(k) (k=1, 2.…, p).     

Analyzing the identifiability of linear dynamic models with parameter space separators

We take a new approach to studying the structural global identifiability of linear dynamic models in the state space using the concept of separators of the parameter space. We offer some criteria for the truth of separators which are based on specially constructed matrices, thereby avoiding the laborious analytic solution of a systemof nonlinear algebraic equations. Some examples are given that illustrate applications of the proposed approach.     

Estimation in partial linear EV models with replicated observations

The aim of this work is to construct the parameter estimators in the partial linear errors-in-variables (EV) models and explore their asymptotic properties. Unlike other related references, the assumption of known error covariance matrix is removed when the sample can be repeatedly drawn at each designed point from the model. The estimators of interested regression parameters, and the model error variance, as well as the non-parametric function, are constructed. Under some regular conditions, all of the estimators prove strongly consistent. Meanwhile, the asymptotic normality for the estimator of regression parameter is also presented. A simulation study is reported to illustrate our asymptotic results.     

Estimation of high-dimensional linear factor models with grouped variables

We introduce a generalization of the approximate factor model that divides the observable variables into groups, allows for arbitrarily strong cross-correlation between the disturbance terms of variables that belong to the same group, and for weak correlation between the disturbances of variables that belong to different groups. We call this model the Grouped Variable Approximate Factor Model. We establish identification, propose an estimation approach based on instrumental variable conditions that hold in the limit, and prove consistency in a dual limit framework. Monte Carlo simulations are used to investigate the performance of the estimator, and the techniques are applied to an analysis of industrial output in the US.     

Estimation of partial linear error-in-response models with validation data

In this paper, an estimation theory in partial linear model is developed when there is measurement error in the response and when validation data are available. A semiparametric method with the primary data is used to define two estimators for both the regression parameter and the nonparametric part using the least squares criterion with the help of validation data. The proposed estimators of the parameter are proved to be strongly consistent and asymptotically normaal, and the estimators of the nonparametric part are also proved to be strongly consistent and weakly consistent with an optimal convergent rate. Then, the two estimators of the parameter are compared based on their empirical performances. Supported by NNSF of China (No. 10231030, No. 10241001) and a grant to the author for his excellent Ph.D. dissertation work in China.     

Estimation in partially linear models with missing responses at random

A partially linear model is considered when the responses are missing at random. Imputation, semiparametric regression surrogate and inverse marginal probability weighted approaches are developed to estimate the regression coefficients and the nonparametric function, respectively. All the proposed estimators for the regression coefficients are shown to be asymptotically normal, and the estimators for the nonparametric function are proved to converge at an optimal rate. A simulation study is conducted to compare the finite sample behavior of the proposed estimators.     

Stabilization of some systems with two linear delays

We obtain methods for stabilizing systems with two linear delays and systems with two linear delays and one constant delay. We present a numerical example in which a stabilizing control is constructed.     

Adaptive anti-synchronization of two identical and different hyperchaotic systems with uncertain parameters

This paper brings attention to hyperchaos anti-synchronization between two identical and different hyperchaotic systems by using adaptive control. The sufficient conditions for achieving the anti-synchronization of two hyperchaotic systems are derived based on Lyapunov stability theory. An adaptive control law and a parameter update rule for unknown parameters are introduced such that the hyperchaotic Chen system is controlled to be the hyperchaotic Lü system. Theoretical analysis and numerical simulations are shown to verify the results.     

Continuum eigenmodes in some linear stellar models

We apply parallel approaches in the study of continuous spectra to adiabatic stellar models. We seek continuum eigenmodes for the LAWE formulated as both finite difference and linear differential equations. In particular, we apply methods of Jacobi matrices and methods of subordinancy theory in these respective formulations. We find certain pressure-density conditions which admit positive-measured sets of continuous oscillation spectra under plausible conditions on density and pressure. We arrive at results of unbounded oscillations and computational or, perhaps, dynamic instability.     

Estimation of the parameters for unstable AR models

ＥＳＴＩＭＡＴＩＯＮＯＦＴＨＥＰＡＲＡＭＥＴＥＲＳＦＯＲＵＮＳＴＡＢＬＥＡＲＭＯＤＥＬＳＡＮＨｏＮＧＺＨＩ（安鸿志）（ＩｎｓｔｉｔｕｔｅｏｆＡｐｐｌｉｅｄＭａｔｈｅｍａｔｉｃｓ,ｔｈｅＣｈｉｎｅｓｅＡｃａｄｅｍｙｏｆＳｃｉｅｎｃｅ,Ｂｅｉｊｉｎｇ１０...     

Comparisons among some estimators in misspecified linear models with multicollinearity

In this paper we deal with comparisons among several estimators available in situations of multicollinearity (e.g., the r-k class estimator proposed by Baye and Parker, the ordinary ridge regression (ORR) estimator, the principal components regression (PCR) estimator and also the ordinary least squares (OLS) estimator) for a misspecified linear model where misspecification is due to omission of some relevant explanatory variables. These comparisons are made in terms of the mean square error (mse) of the estimators of regression coefficients as well as of the predictor of the conditional mean of the dependent variable. It is found that under the same conditions as in the true model, the superiority of the r-k class estimator over the ORR, PCR and OLS estimators and those of the ORR and PCR estimators over the OLS estimator remain unchanged in the misspecified model. Only in the case of comparison between the ORR and PCR estimators, no definite conclusion regarding the mse dominance of one over the other in the misspecified model can be drawn.     

QUADRATIC ESTIMATORS OF QUADRATIC FUNCTIONS WITH PARAMETERS IN NORMAL LINEAR MODELS

ＱＵＡＤＲＡＴＩＣＥＳＴＩＭＡＴＯＲＳＯＦＱＵＡＤＲＡＴＩＣＦＵＮＣＴＩＯＮＳＷＩＴＨＰＡＲＡＭＥＴＥＲＳＩＮＮＯＲＭＡＬＬＩＮＥＡＲＭＯＤＥＬＳ￥ＷＵＱＩＧＵＡＮＧ（吴启光）（ＩｎｓｔｉｔｕｔｅｏｆＳｙｓｔｅｍｅＳｃｉｅｎｃｅ,ｔｈｅＣｈｉｎｅｓｅ...     

A new estimate of the parameters in linear mixed models

In linear mixed models, there are two kinds of unknown parameters: one is the fixed effect, the other is the variance component. In this paper, new estimates of these parameters, called the spectral decomposition estimates, are proposed, Some important statistical properties of the new estimates are established, in particular the linearity of the estimates of the fixed effects with many statistical optimalities. A new method is applied to two important models which are used in economics, finance, and mechanical fields. All estimates obtained have good statistical and practical meaning.     

Classroom note: Estimation of the cross-product of two mean vectors

The problem of estimating the cross-product of two mean vectors in three-dimensional Euclidian space is considered. Two ‘natural’ estimators are developed, both of which turn out to be biased. A third, unbiased estimator, resulting from a jackknife procedure, is also investigated. It is shown that, under normality, the latter is best among all the unbiased estimators of this quantity.     

Regular planar graphs with faces of only two types and shortness parameters

We construct a small non-Hamiltonian 3-connected trivalent planar graph whose faces are all 4-gons or 7-gons and show that the shortness coefficient of the class of such graphs is less than one. Then, by transforming non-Hamiltonian trivalent graphs into regular graphs of valency four or five, we obtain our main results, as follows. We show first that the class of 3-connected r-valent planar graphs whose faces are of only two types, triangles and q-gons, contains non-Hamiltonian members and has a shortness exponent less than one when r = 4, for all q ≧ 12. Under the extra restriction that, among graphs of connectivity three, only those with maximum cyclic edge-connectivity are to be considered, we prove the same result also when r = 4, for q = 20, and when r = 5, for all q ≧ 14 except multiples of three.