Merging Gini s Indices under Quadratic Loss

Size: px
Start display at page:

Download "Merging Gini s Indices under Quadratic Loss"

Transcription

1 Merging Gini s Indices under Quadratic Loss S. E. Ahmed, A. Hussein Department of Mathematics and Statistics University of Windsor Windsor, Ontario, CANADA N9B 3P4 ahmed@uwindsor.ca M. N. Goria Universita Trento Trento, Italy February 11,

2 Abstract The Gini index is perhaps one of the most used indicators of economic and social condition. This article develops simultaneous estimation strategies of the Gini indices when samples are taken from several sources. We consider a basis for optimally combining various data sets. In a multi-sample scenario, we demonstrate that a shrinkage-type estimator has, under quadratic loss, a superior risk performance relative to the conventional estimators. Asymptotic distributional quadratic biases and risks of the proposed estimators are derived and compared with benchmark estimators. Key Words and Phrases: Gini index, Gini mean difference, restricted estimation, shrinkage-type estimation, local alternatives, quadratic loss function, asymptotic distributional quadratic risk, retrospective sampling. 1 INTRODUCTION The Gini index, proposed by Gini in 1912, of income or resource inequality is a measure of the degree to which a population shares that resource unequally. It is based on the statistical notion known in the literature as the mean difference of a population. The research on the Gini index has been developed remarkably and extended into various directions as evidenced by the bibliographies of Xu (2004), Yitzhaki (1998). Further, a substantial literature has been devoted to the construction of indices of economic inequality that are consistent with axiomatic systems of fairness. We refer to Anand (1983), Chakravarty (1990), Sen (1997) and Atkinson and Bourguignon (2000) for comprehensive surveys on the measures of inequality including Gini index. In order to define Gini s mean difference, suppose that X 1,, X n re independent and identically distributed (iid) with cumulative distribution function (cdf) F. We shall assume that F, instead of being restricted to a 2

3 parametric family, is completely unknown, subject only to some very general conditions such as continuity or existence of moments. The parameter θ = θ(f ) to be estimated is a real valued function defined over this nonparametric class F. The population Gini index is defined as γ = 2µ, (1) where is the population Gini s mean difference and is defined as The estimated Gini index is defined as = E X 2 X 1. (2) ˆγ = ˆ 2ˆµ, (3) where ˆµ and ˆ are consistent estimators of µ and, respectively. Further, an unbiased estimator of is given by ˆ = ( 1 ) n 2 i<j X j X i = 1 n n X n 2 j X i. (4) i=1 j=1 For large n, n( ˆ ) follows a normal distribution with finite variance (see Lehmann (1999) and Lee (1990)). 1968, 1970 and references therein) can be written as, ˆγ = 1 2n 2 x The Gini index estimator (see David, X j X i. (5) i=1 j=1 In the present investigation, the simultaneous estimation of Gini indices is considered in a multi-sample situation. Suppose that s independent random samples are obtained either from populations having similar characteristics or at different time points. Thus, we are interested in the analysis of sequences of the data sets collected in separate studies over time or space of the same phenomenon. Data sets obtained in this fashion is at times referred to as meta analysis. 3

4 Let γ l = l /(2µ l ) for l = 1, 2,, s and suppose there are s independent retrospective samples of size n 1, n 2,, n s acquired from these s populations. Denote the observed data by X li, i = 1, 2,, n l with distribution functions F l (x). The population parameters may differ from sample to sample due to a variety of reasons. We define Gini s mean difference parameter-vector as = ( 1, 2,, s ) and its covariance structure as Σ = σ 2 l I (s s), where I is an identity matrix. The population Gini index of each component is γ l = l 2ˆµ l, (6) where, l = E X lj X li. (7) If l and µ l are unknown, then the conventional estimator of ( l, µ l ) is ( ˆ l, ˆµ l ) where ˆ l = 1 ( nl 2 ) i<j Thus, the estimated componentwise Gini index is X lj X li, ˆµ l = 1 n l X li. (8) n l i=1 ˆγ l = ˆ l 2ˆµ l, (9) Now, we consider the estimation problem when the homogeneity of the indices is suspected. In other words, when the hypothesis H 0 : γ 1 = γ 2 = = γ l = γ 0 (unknown) (10) is thought to be true. For the sake of brevity, assume that under such equality of the Gini indices, the distributions of the X jl may be identical, i.e., F l F for all l. Our objective in this article is to consider the problem of simultaneous estimation of under the homogeneity of population indices. However, the 4

5 information in relation (10), regarding the homogeneity, is rather uncertain, and hence must be treated as uncertain prior information (UPI). Now, the question we would like to address in this investigation is: how to incorporate the given UPI in the estimation process? In other words, how to combine the UPI regarding the parameters with the available sample data. Our goal is to develop natural adaptive estimation methods that are free of subjective choices and tuning parameters and have superior risk performance under quadratic loss. In the context of the multi-parameter statistical models, we demonstrate a well-defined data-based shrinkage-type Gini index (SGI) estimator that combines estimation problems by shrinking a base (conventional) estimator to a plausible alternative estimator (estimator under restriction). Asymptotic results are demonstrated and the relationship of the SGI estimator to the family of Stein-rule (SR) estimators is discussed. 2 ESTIMATION STRATEGIES Our main interest is estimating the s-dimensional Gini indices parameter vector γ = (γ 1, γ 2,, γ s ). Throughout this paper, the boldface symbols will represent vectors/matrices. 2.1 Estimation Under Full Model For the full model, the conventional or unrestricted estimator (UE) of γ = (γ 1, γ s ), is defined as ˆγ = (ˆγ 1, ˆγ s ), where, ˆγ l = ˆ l /ˆµ l, l = 1, s. The following lemma provides the asymptotic distribution of the unrestricted estimator. Lemma 1 If EX 2 < then nl (ˆγ l γ l ) d N (0, τ 2 l ), (11) 5

6 where d means convergence in distribution, τ 2 l = 1 µ 2 l ( σ1l γ 2 l 2σ 2l γ l + σ 3l ), (12) σ 1l = σ 2l = σ 3l = x<y x<y x<y 2F l (x)[1 F l (y)]dxdy [(2F l (x) 1) + (2F l (y) 1)]F l (x)[1 F l (y)]dxdy 2(2F l (x) 1)(2F l (y) 1)F l (x)[1 F l (y)]dxdy, (13) and F l (.) is the distribution function for the l th population. The lemma can be proved using results in Stigler (1974) as in Ahmed et al. (2005). Also, it can be shown that ˆγ under quadratic loss is asymptotically minimax with constant quadratic risk trace[diag(τ 2 l )]. 2.2 Estimation Under Reduced Model Under the UPI that γ 1 = γ 2 = = γ s, we propose a restricted estimator (RE) of γ as follows ˆγ R = (ˆγ R, ˆγ R ) = ˆγ R 1 s, s ˆγ R = n lˆγ l /n, n = n n s, (14) l=1 where 1 s is an s-dimensional vector of one. We will show that ˆγ R has smaller asymptotic quadratic risk than ˆγ in an interval near the UPI at the expense of poorer performance in the rest of the parameter space induced by the UPI. Not only that, but also the risk function of ˆγ R becomes unbounded as the UPI error grows. If the prior information regarding homogeneity of the parameters is bad in the sense that the UPI error is large, the restricted estimator is inferior to ˆγ. Alternatively, if the information is good, i.e., the UPI error is small, ˆγ R offers a substantial gain over ˆγ. The above insight leads to shrinkage-type estimation when the information is rather suspicious and it is useful to construct a compromised estimator to identify model-estimator uncertainty. 6

7 2.3 The Shrinkage-type Base As one basis for identifying model-estimator uncertainty. Stein (1956) demonstrated the inadmissibility of the traditional maximum likelihood estimator (MLE) when estimating the s-variate normal mean vector θ under quadratic loss. Following this result, James and Stein (1961), Stein (1962), and Baranchik (1964) combined the s-variate MLE ˆθ with s-dimensional fixed null vector, under the normality assumption, as ˆθ S = (1 c/ ˆθ 0 2 )(ˆθ 0), where 0 < c < 2(s 2), and demonstrated that for s > 2 this estimator dominates the MLE. In an orthonormal s-mean context, Lindley (1962) suggested shrinking ˆθ towards the grand-mean estimator and demonstrated the risk dominance the the Stein Estimator. Ahmed and Saleh (1989), Green and Strawderman (1991) and Kim and White (2001) investigated the properties of the Stein-type estimator under various statistical model settings. Ahmed (1999) investigated the estimation problem of survivor functions in s independent sample setting based on reduced and full models, and demonstrated that, under quadratic loss, it yields the asymptotic distributional quadratic risk-dominating estimator ˆθ A = [1 (s 3)/T ](ˆθ ˆθ R ) + ˆθ, s > 3, where ˆθ R is the restricted estimator, and T = n(ˆσ n) 2 1 (ˆθ ˆθ R ) D(ˆθ ˆθ R ), where ˆσ n 2 is a consistent estimator of the nuisance parameter σ 2 and D = Diag(n 1 /n, n s /n), with n = n 1 + +n s. Clearly, this estimator resembles the Stein-rule estimator. Given this base, Ahmed et al. (2001), considered the simultaneous estimation of several intraclass correlation coefficients when independent samples are drawn from s multivariate normal populations and 7

8 provided an expression for the asymptotic risk and bias of the Stein-type estimator, ˆθ A. Now using the Stein-like base, provided by Ahmed (1999), we propose the following shrinkage-type estimator for the parameter vector, γ, as ˆγ = [1 (s 3)/Tn](ˆγ ˆγ R ) + ˆγ, s > 3, where T n = n(ˆτ 2 n) 1 (ˆγ ˆγ R ) D n (ˆγ ˆγ R ), ˆτ 2 n is a consistent estimator of the common value of τ 2 l, and D n = Diag(n l /n) with l = 1,..., s. The estimator ˆγ is in the general form of the Stein rule family of estimators, where shrinkage of the base estimator ˆγ is toward the alternative estimator ˆγ R. It is interesting to note that the proposed strategy is similar in spirit to the Bayesian model-averaging procedures. However, the main difference is that the Bayesian model-averaging procedures are not optimized with respect to any particular loss function. The present investigation is stimulated by prediction offered by Professor Efron in RSS News of January, The empirical Bayes/James-Stein category was the entry in my list least affected by computer developments. It is ripe for a computer-intensive treatment that brings the substantial benefits of James-Stein estimation to bear on complicated, realistic problems. A side benefit may be at least a partial reconciliation between frequentist and Bayesian perspectives as they apply to statistical practice. It may be worth mentioning that this is one of the two areas Professor Efron predicted for continuing research for the early 21st century. Shrinkage and likelihood-based methods continue to play vital roles in statistical inference. These methods provide extremely useful techniques for combining data from various sources. In passing we would like to remark 8

9 that the preliminary test estimation can also be used for tackling the uncertainty. However, making use of Stein-type estimator Sclove et al. (1972) demonstrated the non-optimality of preliminary test estimation as basis for dealing with model uncertainty. Hence, we confine here on Stein-type estimation, however for s < 3 preliminary test estimation may be a useful choice to tackle the estimation uncertainty. A plan for rest of the paper is as follows. We present some useful asymptotic results in Section 3 which form the basis for our study. Further, the expressions for bias and risk of the proposed estimators are given. The discussion on the risk behavior of the proposed estimators are contained in Section 4. Furthermore, some computed risk analyses are presented in the same section. 3 ASYMPTOTIC RESULTS We shall examine the properties of the proposed estimators under asymptotic set up in the light of the following weighted quadratic loss function: L(γ o, γ) = n(γ o γ) G(γ o γ), (15) where γ o is an appropriate estimator of γ and G is a given positive semidefinite matrix. Assume that G(y) = lim n P { n(γ o γ) y}. Then we define the asymptotic distributional quadratic risk (ADQR) by R(γ o, γ) = where G o = yy dg(y). y GydG(y) = trace(gg o ), (16) Further we consider the following contiguous sequence of alternatives to establish the needed asymptotic results: K (n) : γ = γ n, where γ n = γ o + λ n, λ a fixed real vector. (17) 9

10 Note that λ = 0 implies γ = γ o 1 s, so (10) is a particular case of {K (n) }. Based on regularity conditions no more stringent than the typical types of conditions assumed for establishing asymptotic properties of Gini indices, the proposed estimator also achieves similar properties. Noting that n(ˆγ ˆγ) G n(ˆγ ˆγ) = (s 3) 2 T 2 n { n(ˆγ ˆγ R ) G n(ˆγ ˆγ R )} (s 3) 2 {n(ˆγ ˆγ R ) G(ˆγ ˆγ R )} 1 {ch max (GD n 1 )} 2. where ch max (A) is the largest eigen-root of a matrix A. On the other hand, in the case of ˆγ R, n(ˆγ R γ) G(ˆγ R γ) p +, as n, (18) where p means convergence in probability. By virtue of the above the result, the ADQR of ˆγ R, for any γ H o, will approach + for large n. To compare the respective risk-performance of all the proposed estimators, we establish the following lemmas under the local alternatives, which facilitate the derivation of the ADQR of the proposed estimators. Lemma 2 Let U n = n(ˆγ γ o ), alternatives where ( Un V n ) N 2s {( λλ ), V n = n(ˆγ ˆγ R ), then under local ( τ 2 D 1 )} B B B as n, (19) λ = Hλ, H = I s JD, J = 1 s 1 s, D = lim(d n ), τ 2 = lim(ˆτ 2 n) B = τ 2 D 1 H Lemma 3 Let Z n = n(ˆγ R γ o ) then under local alternatives ( Zn V n ) N 2s {( 0λ ), ( τ 2 )} J 0 0 B 10 as n. (20)

11 By virtue of the above lemmas we shall present expressions for the asymptotic distributional bias (ADB) and ADQR of the estimators in the following section. Let Ψ s (x ; Λ) stand for the noncentral chi-square distribution with noncentrality parameter Λ and s degrees of freedom. Further, E (χ 2m s (Λ)) = 0 x m dψ s (x ; Λ). Let us define the asymptotic distributional bias (ADB) of an estimator γ o as B(γ o ) = lim n E{ n(γ o γ}. We present the expressions for the bias of the estimators in the following theorem. Using lemma 2 and 3, the bias of ˆγ is obtained by the same argument in Ahmed (2001) and by direct computations. Next, we establish the following lemma. Lemma 4 where, Λ = (τ 2 ) 1 λ Dλ. B(ˆγ R ) = λ, B(ˆγ ) = (k 3)λ E(χ 2 k+1 (Λ)), However, in an effort to present a clear analysis of various bias functions, first we transform various bias functions in scalar (quadratic) form by defining B (γ o ) = (τ 2 ) 1 [B(γ o )] D[B(γ o )] as the quadratic bias of the estimator γ o of a parameter vector γ. Thus, B (ˆγ R ) = Λ, B (ˆγ ) = (s 3) 2 Λ[E(χ 2 s+1(λ))] 2, The asymptotic bias functions of both estimators depend upon the parameters only through Λ. Hence, we investigate the behavior of the quadratic bias of the proposed estimators in terms of Λ. It is easy to see that the magnitude of bias of ˆγ R increases without a bound and tends to as Λ. The bias of ˆγ starts from 0 at Λ = 0 then increases to a point then decreases 11

12 towards 0, since E(χ 2 ν (Λ)) is a decreasing log-convex function of Λ. Since bias is a component of ADQR, we will discuss the ADQR of the estimators from here onwards. The expressions for ADQR are given in the following theorem. Theorem 1 For large n, and under local alternatives, the ADQRs of the estimators are given by R(ˆγ, γ) = τ 2 trace(gd 1 ), (21) R(ˆγ R, γ) = τ 2 trace(gd 1 ) trace(gc) + Λ G, (22) where Λ G = λ Gλ C = τ 2 (D 1 J), R(ˆγ, γ) = τ 2 trace(gd 1 ) + Λ G (k 3)(k + 1)E(χ 4 s+3(λ)) (k 3)trace(GC){2E(χ 2 s+1(λ)) (s 3)E(χ 4 s+1(λ))}, (23) Proof of (20) - (21) is fairly straightforward, (22) is obtained by using the same argument as in Ahmed (2001) and direct computations, therefore, we avoid the detail of the derivation. 4 RISK ANALYSIS FOR VARIOUS ESTI- MATORS In this section the large sample properties of the proposed estimators are discussed in the light of the quadratic loss function. The ADQR of ˆγ is constant (independent of Λ) with the value trace(gd 1 ), while the risk of ˆγ R becomes unbounded as the hypothesis error grows crossing the risk of ˆγ. Furthermore, we note that R(ˆγ R ; γ) R(ˆγ; γ) if Λ trace(gc). 12

13 Thus, ˆγ R dominates ˆγ in the interval [0, trace(gc)). Clearly, when Λ moves away from the origin beyond the value trace(gc), the ADQR of ˆγ R increases without a bound. We now turn to investigate the comparative statistical properties of the shrinkage-type estimator. First we compare it with ˆγ when Λ = 0. R(ˆγ; γ) R(ˆγ ; γ) = trace(gc)(s 3)E{2χ 2 s+1(0) (s 3)χ 4 s+1(0)} is a positive quantity. Hence, we conclude that the Stein-type estimator dominates ˆγ at this parametric value. Also, the maximum risk gain of ˆγ over ˆγ is achieved at the same point. To examine the risk behavior of ˆγ when Λ > 0, we characterize a class of positive semi-definite matrices G D = { trace(gd 1 ) ch max (GD 1 ) s + 1 } 2 where ch max (.) means the largest eigenvalue of (.). (24) In order to provide a meaningful comparison of the various estimators, we state the following theorem from linear algebra. Theorem 2 (Courant Theorem) If A and B are two positive semi-definite matrices with B nonsingular, both of order (s s), then ch min (AB 1 ) x Ax x Bx ch max(ab 1 ) where ch min ( ) and ch max ( ) mean the smallest and largest eigenvalues of ( ), respectively, and x is a column vector of order (s 1). We note that the above lower and upper bounds are equal to the infimum and supremum, respectively, of the ratio x Ax for x 0. Also, for B = I, x Bx the ratio is known as Rayleigh quotient for matrix A. 13

14 As a consequence of the Courant theorem, we have λ min (GD 1 ) λ Gλ λ Dλ λ max(gd 1 ), for λ 0 and G G D. Thus, under the class of matrices defined in relation (24) we conclude that for all λ, R(ˆγ ; γ) R(ˆγ; γ) where strict inequality holds for some λ. It clearly indicates the asymptotic inadmissibility of ˆγ under local alternatives relative to ˆγ. The risk of ˆγ begins with an initial value of 3 and increases monotonically towards trace(gd 1 ) as the value of Λ moves away from 0. The risk of ˆγ is uniformly smaller than ˆγ, where the upper limit is attained when λ. The result is valid as long the expectation in risk expression exists, which is the case whenever s 4. Next, R(ˆγ, γ) R(ˆγ R, γ) = trace(gc) s 3 s 1 trace(gd 1 ) > 0. (25) Therefore, the ADQR of ˆγ R is smaller than the ADQR of ˆγ when Λ = 0. Alternatively, when Λ departs from the initial value 0, in turn the value of E(χ 4 s+1(λ)) decreases, so ˆγ has smaller ADQR than ˆγ R. ˆγ dominates ˆγ in the rest of the parameter space. Hence, under local alternatives none of ˆγ and ˆγ R is asymptotically better than the other. To our knowledge, no estimator exists in the class that can outperform the optimality of the estimator based on reduced model (if true) in the entire parameter space. The present investigation reaffirms this unique characteristic of ˆγ R. It is noted that the risk of all the estimators depend on the matrices G and D. In order to facilitate numerical computation of the ADQR functions, we consider the particular case G = (τ 2 ) 1 D and obtain the value of risk expressions on a digital computer. With this substitution, we get R(ˆγ, γ) = s, R(ˆγ R, γ) = 1 + Λ and R(ˆγ, γ) = s + Λ(s 3)(s + 1)E(χ 4 s+3(λ)) (s 1)(s 3){2E(χ 2 s+1(λ)) (s 3)E(χ 4 s+1(λ))}, 14 (26)

15 We have plotted risk functions versus Λ at selected values of s. The graphical results in Figures 1 reinforce our theoretical findings that ˆγ dominates ˆγ. 5 CONCLUDING NOTE In this paper, we discussed estimation strategies for multi-sample Gini indices based on full and reduced models in the presence of uncertain prior information. The performance of the restricted estimator (i.e., under the UPI) heavily depends on the quality of non-sample information. However, one is seldom sure of the reliability of this information. We have presented shrinkage-type estimator for this multi-sample Gini indices. We find that ˆγ is relatively more efficient than ˆγ in the entire parameter space. It was also noted that ˆγ and can only be used for s > 3. It is worth mentioning that the assumption of homogeneity of population distributions under the hypothesis of equal Gini indices can be relaxed in which case the restricted estimator will be weighted by the asymptotic variances of the vector of Gini indices. References Ahmed, S. E. (1999). Simultaneous estimation of survivor functions in exponential lifetime models. Journal of Statistical Computation and Simulation 63, Ahmed, S. E., A. K. Gupta, S. M. Khan and C. J. Nicol (2001). Simultaneous estimation of several intraclass correlation coefficients. Annals of the Institute of Statistical Mathematics 53(2), Ahmed, S. E. and A. K. Md. E. Saleh (1989). Pooling multivariate data. Journal of Statistical Computation and Simulation 31,

16 Risk s=4 UE RE JS Risk s= Lambda Lambda Risk s=12 Risk s= Lambda Lambda Figure 1: Quadratic risk plotted as function of noncentrality parameter, Λ for the three estimators described above for various number of populations, s. 16

17 Ahmed, S.E. (2001). Shrinkage estimation of regression coefficients from censored data with multiple observations. Empirical Bayes and Likelihood Inference, Lecture Notes in Statistics, Editors: S.E. Ahmed and N. Reid. Springer-Verlag. Ahmed, S.E., A.A. Hussein and R. Ghori (2005). Gini mean difference and its applications in robust estimation.. Tecnical Report, University of Windsor, Canada. Anand, S. (1983). Inequality and poverty in Malaysia: Measurement and Decomposition. Oxford University Press. Atkinson, A.B. and F. Bourguignon (2000). Introduction: Income distribution and Economics, Handbook of Income Distribution. Elsevier. Baranchik, A.M. (1964). Multiple regression ans estimation of the mean of a multivariate normal distribution.. Technical Report 51, Stanford University, Dept. of Statistics. Chakravarty, S. R. (1990). Ethical Social Index Numbers. Springer. David, H.A. (1968). Gini s mean difference rediscovered. Biometrika 55, David, H.A. (1970). Order Statistics. Wiley. Green, Edwin J. and William E. Strawderman (1991). A James-Stein type estimator for combining unbiased and possibly biased estimators. Journal of the American Statistical Association 86, James, W. and C. Stein (1961). Estimation with quadratic loss. Proceeding of the Fourth Berkeley Symposium On Mathematical Statistics and Probability.. University of California Press, Berkeley, CA. 17

18 Kim, Tae-Hwan and Halbert White (2001). James-Stein-type estimators in large samples with application to the least absolute deviations estimator. Journal of the American Statistical Association 96(454), Lee, A. J. (1990). U-Statistics. Marcel Dekker, New York. Lehmann, E. L. (1999). Elements of Large-sample Theory. Springer-Verlag Inc. Lindley, D.V. (1962). Discussion of professor stein s paper. Journal of the Royal Statistical Society, B 24, Sclove, S. L., C. Morris and R. Radhakrishnan (1972). Non optimality of preliminary test estimation for the multinormal mean.. The Annals of Mathematical Statistics 43, Sen, A.K. (1997). On Economic Inequality.. Oxford:Clarendon Press. Stein, C. (1956). Inadmissibility of the usual estimator of the mean of a multivariate normal distribution. Proceeding of the Fourth Berkeley Symposium On Mathematical Statistics and Probability.. University of California Press, Berkeley, CA. Stein, C. (1962). Confidence sets for the mean of a multivariate normal distribution.. Journal of the Royal Statistical Society, B 24, Stigler, S.M. (1974). Linear functions of order statistics with smooth weight functions. Ann. Statist. 2, Xu, K. (2004). How has the literature on Gini s index evolved in the past 80 years?. Technical Report, Dalhousie University. Yitzhaki, S. (1998). More than a dozen alternative ways of speling gini.. Research in economic inequality 8,

Journal of Statistical Research 2007, Vol. 41, No. 1, pp Bangladesh

Journal of Statistical Research 2007, Vol. 41, No. 1, pp Bangladesh Journal of Statistical Research 007, Vol. 4, No., pp. 5 Bangladesh ISSN 056-4 X ESTIMATION OF AUTOREGRESSIVE COEFFICIENT IN AN ARMA(, ) MODEL WITH VAGUE INFORMATION ON THE MA COMPONENT M. Ould Haye School

More information

Carl N. Morris. University of Texas

Carl N. Morris. University of Texas EMPIRICAL BAYES: A FREQUENCY-BAYES COMPROMISE Carl N. Morris University of Texas Empirical Bayes research has expanded significantly since the ground-breaking paper (1956) of Herbert Robbins, and its province

More information

Shrinkage Estimation of the Slope Parameters of two Parallel Regression Lines Under Uncertain Prior Information

Shrinkage Estimation of the Slope Parameters of two Parallel Regression Lines Under Uncertain Prior Information Shrinkage Estimation of the Slope Parameters of two Parallel Regression Lines Under Uncertain Prior Information Shahjahan Khan Department of Mathematics & Computing University of Southern Queensland Toowoomba,

More information

UC Berkeley CUDARE Working Papers

UC Berkeley CUDARE Working Papers UC Berkeley CUDARE Working Papers Title A Semi-Parametric Basis for Combining Estimation Problems Under Quadratic Loss Permalink https://escholarship.org/uc/item/8z5jw3 Authors Judge, George G. Mittelhammer,

More information

Testing Statistical Hypotheses

Testing Statistical Hypotheses E.L. Lehmann Joseph P. Romano Testing Statistical Hypotheses Third Edition 4y Springer Preface vii I Small-Sample Theory 1 1 The General Decision Problem 3 1.1 Statistical Inference and Statistical Decisions

More information

Fixed Effects, Invariance, and Spatial Variation in Intergenerational Mobility

Fixed Effects, Invariance, and Spatial Variation in Intergenerational Mobility American Economic Review: Papers & Proceedings 2016, 106(5): 400 404 http://dx.doi.org/10.1257/aer.p20161082 Fixed Effects, Invariance, and Spatial Variation in Intergenerational Mobility By Gary Chamberlain*

More information

IEOR 165 Lecture 7 1 Bias-Variance Tradeoff

IEOR 165 Lecture 7 1 Bias-Variance Tradeoff IEOR 165 Lecture 7 Bias-Variance Tradeoff 1 Bias-Variance Tradeoff Consider the case of parametric regression with β R, and suppose we would like to analyze the error of the estimate ˆβ in comparison to

More information

LECTURES IN ECONOMETRIC THEORY. John S. Chipman. University of Minnesota

LECTURES IN ECONOMETRIC THEORY. John S. Chipman. University of Minnesota LCTURS IN CONOMTRIC THORY John S. Chipman University of Minnesota Chapter 5. Minimax estimation 5.. Stein s theorem and the regression model. It was pointed out in Chapter 2, section 2.2, that if no a

More information

Testing Equality of Two Intercepts for the Parallel Regression Model with Non-sample Prior Information

Testing Equality of Two Intercepts for the Parallel Regression Model with Non-sample Prior Information Testing Equality of Two Intercepts for the Parallel Regression Model with Non-sample Prior Information Budi Pratikno 1 and Shahjahan Khan 2 1 Department of Mathematics and Natural Science Jenderal Soedirman

More information

g-priors for Linear Regression

g-priors for Linear Regression Stat60: Bayesian Modeling and Inference Lecture Date: March 15, 010 g-priors for Linear Regression Lecturer: Michael I. Jordan Scribe: Andrew H. Chan 1 Linear regression and g-priors In the last lecture,

More information

Testing Statistical Hypotheses

Testing Statistical Hypotheses E.L. Lehmann Joseph P. Romano, 02LEu1 ttd ~Lt~S Testing Statistical Hypotheses Third Edition With 6 Illustrations ~Springer 2 The Probability Background 28 2.1 Probability and Measure 28 2.2 Integration.........

More information

ORDER RESTRICTED STATISTICAL INFERENCE ON LORENZ CURVES OF PARETO DISTRIBUTIONS. Myongsik Oh. 1. Introduction

ORDER RESTRICTED STATISTICAL INFERENCE ON LORENZ CURVES OF PARETO DISTRIBUTIONS. Myongsik Oh. 1. Introduction J. Appl. Math & Computing Vol. 13(2003), No. 1-2, pp. 457-470 ORDER RESTRICTED STATISTICAL INFERENCE ON LORENZ CURVES OF PARETO DISTRIBUTIONS Myongsik Oh Abstract. The comparison of two or more Lorenz

More information

Statistical Inference with Monotone Incomplete Multivariate Normal Data

Statistical Inference with Monotone Incomplete Multivariate Normal Data Statistical Inference with Monotone Incomplete Multivariate Normal Data p. 1/4 Statistical Inference with Monotone Incomplete Multivariate Normal Data This talk is based on joint work with my wonderful

More information

Testing Some Covariance Structures under a Growth Curve Model in High Dimension

Testing Some Covariance Structures under a Growth Curve Model in High Dimension Department of Mathematics Testing Some Covariance Structures under a Growth Curve Model in High Dimension Muni S. Srivastava and Martin Singull LiTH-MAT-R--2015/03--SE Department of Mathematics Linköping

More information

Parametric Techniques Lecture 3

Parametric Techniques Lecture 3 Parametric Techniques Lecture 3 Jason Corso SUNY at Buffalo 22 January 2009 J. Corso (SUNY at Buffalo) Parametric Techniques Lecture 3 22 January 2009 1 / 39 Introduction In Lecture 2, we learned how to

More information

STA 732: Inference. Notes 10. Parameter Estimation from a Decision Theoretic Angle. Other resources

STA 732: Inference. Notes 10. Parameter Estimation from a Decision Theoretic Angle. Other resources STA 732: Inference Notes 10. Parameter Estimation from a Decision Theoretic Angle Other resources 1 Statistical rules, loss and risk We saw that a major focus of classical statistics is comparing various

More information

Estimation of parametric functions in Downton s bivariate exponential distribution

Estimation of parametric functions in Downton s bivariate exponential distribution Estimation of parametric functions in Downton s bivariate exponential distribution George Iliopoulos Department of Mathematics University of the Aegean 83200 Karlovasi, Samos, Greece e-mail: geh@aegean.gr

More information

Lecture 7 Introduction to Statistical Decision Theory

Lecture 7 Introduction to Statistical Decision Theory Lecture 7 Introduction to Statistical Decision Theory I-Hsiang Wang Department of Electrical Engineering National Taiwan University ihwang@ntu.edu.tw December 20, 2016 1 / 55 I-Hsiang Wang IT Lecture 7

More information

Studentization and Prediction in a Multivariate Normal Setting

Studentization and Prediction in a Multivariate Normal Setting Studentization and Prediction in a Multivariate Normal Setting Morris L. Eaton University of Minnesota School of Statistics 33 Ford Hall 4 Church Street S.E. Minneapolis, MN 55455 USA eaton@stat.umn.edu

More information

Stat 5101 Lecture Notes

Stat 5101 Lecture Notes Stat 5101 Lecture Notes Charles J. Geyer Copyright 1998, 1999, 2000, 2001 by Charles J. Geyer May 7, 2001 ii Stat 5101 (Geyer) Course Notes Contents 1 Random Variables and Change of Variables 1 1.1 Random

More information

Parametric Techniques

Parametric Techniques Parametric Techniques Jason J. Corso SUNY at Buffalo J. Corso (SUNY at Buffalo) Parametric Techniques 1 / 39 Introduction When covering Bayesian Decision Theory, we assumed the full probabilistic structure

More information

COMPARISON OF FIVE TESTS FOR THE COMMON MEAN OF SEVERAL MULTIVARIATE NORMAL POPULATIONS

COMPARISON OF FIVE TESTS FOR THE COMMON MEAN OF SEVERAL MULTIVARIATE NORMAL POPULATIONS Communications in Statistics - Simulation and Computation 33 (2004) 431-446 COMPARISON OF FIVE TESTS FOR THE COMMON MEAN OF SEVERAL MULTIVARIATE NORMAL POPULATIONS K. Krishnamoorthy and Yong Lu Department

More information

Research Article A Nonparametric Two-Sample Wald Test of Equality of Variances

Research Article A Nonparametric Two-Sample Wald Test of Equality of Variances Advances in Decision Sciences Volume 211, Article ID 74858, 8 pages doi:1.1155/211/74858 Research Article A Nonparametric Two-Sample Wald Test of Equality of Variances David Allingham 1 andj.c.w.rayner

More information

Some History of Optimality

Some History of Optimality IMS Lecture Notes- Monograph Series Optimality: The Third Erich L. Lehmann Symposium Vol. 57 (2009) 11-17 @ Institute of Mathematical Statistics, 2009 DOl: 10.1214/09-LNMS5703 Erich L. Lehmann University

More information

Statistics: Learning models from data

Statistics: Learning models from data DS-GA 1002 Lecture notes 5 October 19, 2015 Statistics: Learning models from data Learning models from data that are assumed to be generated probabilistically from a certain unknown distribution is a crucial

More information

Multivariate Distributions

Multivariate Distributions IEOR E4602: Quantitative Risk Management Spring 2016 c 2016 by Martin Haugh Multivariate Distributions We will study multivariate distributions in these notes, focusing 1 in particular on multivariate

More information

To appear in The American Statistician vol. 61 (2007) pp

To appear in The American Statistician vol. 61 (2007) pp How Can the Score Test Be Inconsistent? David A Freedman ABSTRACT: The score test can be inconsistent because at the MLE under the null hypothesis the observed information matrix generates negative variance

More information

ESTIMATORS FOR GAUSSIAN MODELS HAVING A BLOCK-WISE STRUCTURE

ESTIMATORS FOR GAUSSIAN MODELS HAVING A BLOCK-WISE STRUCTURE Statistica Sinica 9 2009, 885-903 ESTIMATORS FOR GAUSSIAN MODELS HAVING A BLOCK-WISE STRUCTURE Lawrence D. Brown and Linda H. Zhao University of Pennsylvania Abstract: Many multivariate Gaussian models

More information

A General Overview of Parametric Estimation and Inference Techniques.

A General Overview of Parametric Estimation and Inference Techniques. A General Overview of Parametric Estimation and Inference Techniques. Moulinath Banerjee University of Michigan September 11, 2012 The object of statistical inference is to glean information about an underlying

More information

STATS 200: Introduction to Statistical Inference. Lecture 29: Course review

STATS 200: Introduction to Statistical Inference. Lecture 29: Course review STATS 200: Introduction to Statistical Inference Lecture 29: Course review Course review We started in Lecture 1 with a fundamental assumption: Data is a realization of a random process. The goal throughout

More information

Inferences on a Normal Covariance Matrix and Generalized Variance with Monotone Missing Data

Inferences on a Normal Covariance Matrix and Generalized Variance with Monotone Missing Data Journal of Multivariate Analysis 78, 6282 (2001) doi:10.1006jmva.2000.1939, available online at http:www.idealibrary.com on Inferences on a Normal Covariance Matrix and Generalized Variance with Monotone

More information

HAPPY BIRTHDAY CHARLES

HAPPY BIRTHDAY CHARLES HAPPY BIRTHDAY CHARLES MY TALK IS TITLED: Charles Stein's Research Involving Fixed Sample Optimality, Apart from Multivariate Normal Minimax Shrinkage [aka: Everything Else that Charles Wrote] Lawrence

More information

This model of the conditional expectation is linear in the parameters. A more practical and relaxed attitude towards linear regression is to say that

This model of the conditional expectation is linear in the parameters. A more practical and relaxed attitude towards linear regression is to say that Linear Regression For (X, Y ) a pair of random variables with values in R p R we assume that E(Y X) = β 0 + with β R p+1. p X j β j = (1, X T )β j=1 This model of the conditional expectation is linear

More information

Lecture 20 May 18, Empirical Bayes Interpretation [Efron & Morris 1973]

Lecture 20 May 18, Empirical Bayes Interpretation [Efron & Morris 1973] Stats 300C: Theory of Statistics Spring 2018 Lecture 20 May 18, 2018 Prof. Emmanuel Candes Scribe: Will Fithian and E. Candes 1 Outline 1. Stein s Phenomenon 2. Empirical Bayes Interpretation of James-Stein

More information

EMPIRICAL ENVELOPE MLE AND LR TESTS. Mai Zhou University of Kentucky

EMPIRICAL ENVELOPE MLE AND LR TESTS. Mai Zhou University of Kentucky EMPIRICAL ENVELOPE MLE AND LR TESTS Mai Zhou University of Kentucky Summary We study in this paper some nonparametric inference problems where the nonparametric maximum likelihood estimator (NPMLE) are

More information

A Test for Order Restriction of Several Multivariate Normal Mean Vectors against all Alternatives when the Covariance Matrices are Unknown but Common

A Test for Order Restriction of Several Multivariate Normal Mean Vectors against all Alternatives when the Covariance Matrices are Unknown but Common Journal of Statistical Theory and Applications Volume 11, Number 1, 2012, pp. 23-45 ISSN 1538-7887 A Test for Order Restriction of Several Multivariate Normal Mean Vectors against all Alternatives when

More information

Analysis of the AIC Statistic for Optimal Detection of Small Changes in Dynamic Systems

Analysis of the AIC Statistic for Optimal Detection of Small Changes in Dynamic Systems Analysis of the AIC Statistic for Optimal Detection of Small Changes in Dynamic Systems Jeremy S. Conner and Dale E. Seborg Department of Chemical Engineering University of California, Santa Barbara, CA

More information

1.1 Basis of Statistical Decision Theory

1.1 Basis of Statistical Decision Theory ECE598: Information-theoretic methods in high-dimensional statistics Spring 2016 Lecture 1: Introduction Lecturer: Yihong Wu Scribe: AmirEmad Ghassami, Jan 21, 2016 [Ed. Jan 31] Outline: Introduction of

More information

Hypothesis Testing. 1 Definitions of test statistics. CB: chapter 8; section 10.3

Hypothesis Testing. 1 Definitions of test statistics. CB: chapter 8; section 10.3 Hypothesis Testing CB: chapter 8; section 0.3 Hypothesis: statement about an unknown population parameter Examples: The average age of males in Sweden is 7. (statement about population mean) The lowest

More information

Central Limit Theorem ( 5.3)

Central Limit Theorem ( 5.3) Central Limit Theorem ( 5.3) Let X 1, X 2,... be a sequence of independent random variables, each having n mean µ and variance σ 2. Then the distribution of the partial sum S n = X i i=1 becomes approximately

More information

Review of Classical Least Squares. James L. Powell Department of Economics University of California, Berkeley

Review of Classical Least Squares. James L. Powell Department of Economics University of California, Berkeley Review of Classical Least Squares James L. Powell Department of Economics University of California, Berkeley The Classical Linear Model The object of least squares regression methods is to model and estimate

More information

Statistical Data Analysis Stat 3: p-values, parameter estimation

Statistical Data Analysis Stat 3: p-values, parameter estimation Statistical Data Analysis Stat 3: p-values, parameter estimation London Postgraduate Lectures on Particle Physics; University of London MSci course PH4515 Glen Cowan Physics Department Royal Holloway,

More information

FULL LIKELIHOOD INFERENCES IN THE COX MODEL

FULL LIKELIHOOD INFERENCES IN THE COX MODEL October 20, 2007 FULL LIKELIHOOD INFERENCES IN THE COX MODEL BY JIAN-JIAN REN 1 AND MAI ZHOU 2 University of Central Florida and University of Kentucky Abstract We use the empirical likelihood approach

More information

14.30 Introduction to Statistical Methods in Economics Spring 2009

14.30 Introduction to Statistical Methods in Economics Spring 2009 MIT OpenCourseWare http://ocw.mit.edu 4.0 Introduction to Statistical Methods in Economics Spring 009 For information about citing these materials or our Terms of Use, visit: http://ocw.mit.edu/terms.

More information

A nonparametric two-sample wald test of equality of variances

A nonparametric two-sample wald test of equality of variances University of Wollongong Research Online Faculty of Informatics - Papers (Archive) Faculty of Engineering and Information Sciences 211 A nonparametric two-sample wald test of equality of variances David

More information

MAXIMUM LIKELIHOOD IN GENERALIZED FIXED SCORE FACTOR ANALYSIS 1. INTRODUCTION

MAXIMUM LIKELIHOOD IN GENERALIZED FIXED SCORE FACTOR ANALYSIS 1. INTRODUCTION MAXIMUM LIKELIHOOD IN GENERALIZED FIXED SCORE FACTOR ANALYSIS JAN DE LEEUW ABSTRACT. We study the weighted least squares fixed rank approximation problem in which the weight matrices depend on unknown

More information

ON THE FAILURE RATE ESTIMATION OF THE INVERSE GAUSSIAN DISTRIBUTION

ON THE FAILURE RATE ESTIMATION OF THE INVERSE GAUSSIAN DISTRIBUTION ON THE FAILURE RATE ESTIMATION OF THE INVERSE GAUSSIAN DISTRIBUTION ZHENLINYANGandRONNIET.C.LEE Department of Statistics and Applied Probability, National University of Singapore, 3 Science Drive 2, Singapore

More information

Statistical Inference On the High-dimensional Gaussian Covarianc

Statistical Inference On the High-dimensional Gaussian Covarianc Statistical Inference On the High-dimensional Gaussian Covariance Matrix Department of Mathematical Sciences, Clemson University June 6, 2011 Outline Introduction Problem Setup Statistical Inference High-Dimensional

More information

A Very Brief Summary of Statistical Inference, and Examples

A Very Brief Summary of Statistical Inference, and Examples A Very Brief Summary of Statistical Inference, and Examples Trinity Term 2009 Prof. Gesine Reinert Our standard situation is that we have data x = x 1, x 2,..., x n, which we view as realisations of random

More information

Data Mining Stat 588

Data Mining Stat 588 Data Mining Stat 588 Lecture 02: Linear Methods for Regression Department of Statistics & Biostatistics Rutgers University September 13 2011 Regression Problem Quantitative generic output variable Y. Generic

More information

Estimation of the Conditional Variance in Paired Experiments

Estimation of the Conditional Variance in Paired Experiments Estimation of the Conditional Variance in Paired Experiments Alberto Abadie & Guido W. Imbens Harvard University and BER June 008 Abstract In paired randomized experiments units are grouped in pairs, often

More information

Statistical Inference: Estimation and Confidence Intervals Hypothesis Testing

Statistical Inference: Estimation and Confidence Intervals Hypothesis Testing Statistical Inference: Estimation and Confidence Intervals Hypothesis Testing 1 In most statistics problems, we assume that the data have been generated from some unknown probability distribution. We desire

More information

Mean squared error matrix comparison of least aquares and Stein-rule estimators for regression coefficients under non-normal disturbances

Mean squared error matrix comparison of least aquares and Stein-rule estimators for regression coefficients under non-normal disturbances METRON - International Journal of Statistics 2008, vol. LXVI, n. 3, pp. 285-298 SHALABH HELGE TOUTENBURG CHRISTIAN HEUMANN Mean squared error matrix comparison of least aquares and Stein-rule estimators

More information

MSE Performance of the Weighted Average. Estimators Consisting of Shrinkage Estimators

MSE Performance of the Weighted Average. Estimators Consisting of Shrinkage Estimators MSE Performance of the Weighted Average Estimators Consisting of Shrinkage Estimators Akio Namba Kazuhiro Ohtani March 215 Discussion Paper No.1513 GRADUATE SCHOOL OF ECONOMICS KOBE UNIVERSITY ROKKO, KOBE,

More information

Notes, March 4, 2013, R. Dudley Maximum likelihood estimation: actual or supposed

Notes, March 4, 2013, R. Dudley Maximum likelihood estimation: actual or supposed 18.466 Notes, March 4, 2013, R. Dudley Maximum likelihood estimation: actual or supposed 1. MLEs in exponential families Let f(x,θ) for x X and θ Θ be a likelihood function, that is, for present purposes,

More information

Frequentist-Bayesian Model Comparisons: A Simple Example

Frequentist-Bayesian Model Comparisons: A Simple Example Frequentist-Bayesian Model Comparisons: A Simple Example Consider data that consist of a signal y with additive noise: Data vector (N elements): D = y + n The additive noise n has zero mean and diagonal

More information

Size and Shape of Confidence Regions from Extended Empirical Likelihood Tests

Size and Shape of Confidence Regions from Extended Empirical Likelihood Tests Biometrika (2014),,, pp. 1 13 C 2014 Biometrika Trust Printed in Great Britain Size and Shape of Confidence Regions from Extended Empirical Likelihood Tests BY M. ZHOU Department of Statistics, University

More information

Lecture 15. Hypothesis testing in the linear model

Lecture 15. Hypothesis testing in the linear model 14. Lecture 15. Hypothesis testing in the linear model Lecture 15. Hypothesis testing in the linear model 1 (1 1) Preliminary lemma 15. Hypothesis testing in the linear model 15.1. Preliminary lemma Lemma

More information

Peter Hoff Minimax estimation October 31, Motivation and definition. 2 Least favorable prior 3. 3 Least favorable prior sequence 11

Peter Hoff Minimax estimation October 31, Motivation and definition. 2 Least favorable prior 3. 3 Least favorable prior sequence 11 Contents 1 Motivation and definition 1 2 Least favorable prior 3 3 Least favorable prior sequence 11 4 Nonparametric problems 15 5 Minimax and admissibility 18 6 Superefficiency and sparsity 19 Most of

More information

Testing a Normal Covariance Matrix for Small Samples with Monotone Missing Data

Testing a Normal Covariance Matrix for Small Samples with Monotone Missing Data Applied Mathematical Sciences, Vol 3, 009, no 54, 695-70 Testing a Normal Covariance Matrix for Small Samples with Monotone Missing Data Evelina Veleva Rousse University A Kanchev Department of Numerical

More information

Combining Biased and Unbiased Estimators in High Dimensions. (joint work with Ed Green, Rutgers University)

Combining Biased and Unbiased Estimators in High Dimensions. (joint work with Ed Green, Rutgers University) Combining Biased and Unbiased Estimators in High Dimensions Bill Strawderman Rutgers University (joint work with Ed Green, Rutgers University) OUTLINE: I. Introduction II. Some remarks on Shrinkage Estimators

More information

Monitoring Random Start Forward Searches for Multivariate Data

Monitoring Random Start Forward Searches for Multivariate Data Monitoring Random Start Forward Searches for Multivariate Data Anthony C. Atkinson 1, Marco Riani 2, and Andrea Cerioli 2 1 Department of Statistics, London School of Economics London WC2A 2AE, UK, a.c.atkinson@lse.ac.uk

More information

HANDBOOK OF APPLICABLE MATHEMATICS

HANDBOOK OF APPLICABLE MATHEMATICS HANDBOOK OF APPLICABLE MATHEMATICS Chief Editor: Walter Ledermann Volume VI: Statistics PART A Edited by Emlyn Lloyd University of Lancaster A Wiley-Interscience Publication JOHN WILEY & SONS Chichester

More information

Asymptotic Normality under Two-Phase Sampling Designs

Asymptotic Normality under Two-Phase Sampling Designs Asymptotic Normality under Two-Phase Sampling Designs Jiahua Chen and J. N. K. Rao University of Waterloo and University of Carleton Abstract Large sample properties of statistical inferences in the context

More information

Part III. A Decision-Theoretic Approach and Bayesian testing

Part III. A Decision-Theoretic Approach and Bayesian testing Part III A Decision-Theoretic Approach and Bayesian testing 1 Chapter 10 Bayesian Inference as a Decision Problem The decision-theoretic framework starts with the following situation. We would like to

More information

More on nuisance parameters

More on nuisance parameters BS2 Statistical Inference, Lecture 3, Hilary Term 2009 January 30, 2009 Suppose that there is a minimal sufficient statistic T = t(x ) partitioned as T = (S, C) = (s(x ), c(x )) where: C1: the distribution

More information

Gaussian Estimation under Attack Uncertainty

Gaussian Estimation under Attack Uncertainty Gaussian Estimation under Attack Uncertainty Tara Javidi Yonatan Kaspi Himanshu Tyagi Abstract We consider the estimation of a standard Gaussian random variable under an observation attack where an adversary

More information

The properties of L p -GMM estimators

The properties of L p -GMM estimators The properties of L p -GMM estimators Robert de Jong and Chirok Han Michigan State University February 2000 Abstract This paper considers Generalized Method of Moment-type estimators for which a criterion

More information

Lecture 5 September 19

Lecture 5 September 19 IFT 6269: Probabilistic Graphical Models Fall 2016 Lecture 5 September 19 Lecturer: Simon Lacoste-Julien Scribe: Sébastien Lachapelle Disclaimer: These notes have only been lightly proofread. 5.1 Statistical

More information

Generalized Multivariate Rank Type Test Statistics via Spatial U-Quantiles

Generalized Multivariate Rank Type Test Statistics via Spatial U-Quantiles Generalized Multivariate Rank Type Test Statistics via Spatial U-Quantiles Weihua Zhou 1 University of North Carolina at Charlotte and Robert Serfling 2 University of Texas at Dallas Final revision for

More information

FE670 Algorithmic Trading Strategies. Stevens Institute of Technology

FE670 Algorithmic Trading Strategies. Stevens Institute of Technology FE670 Algorithmic Trading Strategies Lecture 8. Robust Portfolio Optimization Steve Yang Stevens Institute of Technology 10/17/2013 Outline 1 Robust Mean-Variance Formulations 2 Uncertain in Expected Return

More information

Gaussian processes. Chuong B. Do (updated by Honglak Lee) November 22, 2008

Gaussian processes. Chuong B. Do (updated by Honglak Lee) November 22, 2008 Gaussian processes Chuong B Do (updated by Honglak Lee) November 22, 2008 Many of the classical machine learning algorithms that we talked about during the first half of this course fit the following pattern:

More information

Principles of Statistics

Principles of Statistics Part II Year 2018 2017 2016 2015 2014 2013 2012 2011 2010 2009 2008 2007 2006 2005 2018 81 Paper 4, Section II 28K Let g : R R be an unknown function, twice continuously differentiable with g (x) M for

More information

Testing for a unit root in an ar(1) model using three and four moment approximations: symmetric distributions

Testing for a unit root in an ar(1) model using three and four moment approximations: symmetric distributions Hong Kong Baptist University HKBU Institutional Repository Department of Economics Journal Articles Department of Economics 1998 Testing for a unit root in an ar(1) model using three and four moment approximations:

More information

Chapter 14 Stein-Rule Estimation

Chapter 14 Stein-Rule Estimation Chapter 14 Stein-Rule Estimation The ordinary least squares estimation of regression coefficients in linear regression model provides the estimators having minimum variance in the class of linear and unbiased

More information

Brief Review on Estimation Theory

Brief Review on Estimation Theory Brief Review on Estimation Theory K. Abed-Meraim ENST PARIS, Signal and Image Processing Dept. abed@tsi.enst.fr This presentation is essentially based on the course BASTA by E. Moulines Brief review on

More information

The Restricted Likelihood Ratio Test at the Boundary in Autoregressive Series

The Restricted Likelihood Ratio Test at the Boundary in Autoregressive Series The Restricted Likelihood Ratio Test at the Boundary in Autoregressive Series Willa W. Chen Rohit S. Deo July 6, 009 Abstract. The restricted likelihood ratio test, RLRT, for the autoregressive coefficient

More information

LECTURE 5 NOTES. n t. t Γ(a)Γ(b) pt+a 1 (1 p) n t+b 1. The marginal density of t is. Γ(t + a)γ(n t + b) Γ(n + a + b)

LECTURE 5 NOTES. n t. t Γ(a)Γ(b) pt+a 1 (1 p) n t+b 1. The marginal density of t is. Γ(t + a)γ(n t + b) Γ(n + a + b) LECTURE 5 NOTES 1. Bayesian point estimators. In the conventional (frequentist) approach to statistical inference, the parameter θ Θ is considered a fixed quantity. In the Bayesian approach, it is considered

More information

ACTEX CAS EXAM 3 STUDY GUIDE FOR MATHEMATICAL STATISTICS

ACTEX CAS EXAM 3 STUDY GUIDE FOR MATHEMATICAL STATISTICS ACTEX CAS EXAM 3 STUDY GUIDE FOR MATHEMATICAL STATISTICS TABLE OF CONTENTS INTRODUCTORY NOTE NOTES AND PROBLEM SETS Section 1 - Point Estimation 1 Problem Set 1 15 Section 2 - Confidence Intervals and

More information

A better way to bootstrap pairs

A better way to bootstrap pairs A better way to bootstrap pairs Emmanuel Flachaire GREQAM - Université de la Méditerranée CORE - Université Catholique de Louvain April 999 Abstract In this paper we are interested in heteroskedastic regression

More information

ASSESSING A VECTOR PARAMETER

ASSESSING A VECTOR PARAMETER SUMMARY ASSESSING A VECTOR PARAMETER By D.A.S. Fraser and N. Reid Department of Statistics, University of Toronto St. George Street, Toronto, Canada M5S 3G3 dfraser@utstat.toronto.edu Some key words. Ancillary;

More information

Physics 403. Segev BenZvi. Parameter Estimation, Correlations, and Error Bars. Department of Physics and Astronomy University of Rochester

Physics 403. Segev BenZvi. Parameter Estimation, Correlations, and Error Bars. Department of Physics and Astronomy University of Rochester Physics 403 Parameter Estimation, Correlations, and Error Bars Segev BenZvi Department of Physics and Astronomy University of Rochester Table of Contents 1 Review of Last Class Best Estimates and Reliability

More information

Preliminaries The bootstrap Bias reduction Hypothesis tests Regression Confidence intervals Time series Final remark. Bootstrap inference

Preliminaries The bootstrap Bias reduction Hypothesis tests Regression Confidence intervals Time series Final remark. Bootstrap inference 1 / 171 Bootstrap inference Francisco Cribari-Neto Departamento de Estatística Universidade Federal de Pernambuco Recife / PE, Brazil email: cribari@gmail.com October 2013 2 / 171 Unpaid advertisement

More information

Testing Algebraic Hypotheses

Testing Algebraic Hypotheses Testing Algebraic Hypotheses Mathias Drton Department of Statistics University of Chicago 1 / 18 Example: Factor analysis Multivariate normal model based on conditional independence given hidden variable:

More information

Analysis of variance and linear contrasts in experimental design with generalized secant hyperbolic distribution

Analysis of variance and linear contrasts in experimental design with generalized secant hyperbolic distribution Journal of Computational and Applied Mathematics 216 (2008) 545 553 www.elsevier.com/locate/cam Analysis of variance and linear contrasts in experimental design with generalized secant hyperbolic distribution

More information

Fall 2017 STAT 532 Homework Peter Hoff. 1. Let P be a probability measure on a collection of sets A.

Fall 2017 STAT 532 Homework Peter Hoff. 1. Let P be a probability measure on a collection of sets A. 1. Let P be a probability measure on a collection of sets A. (a) For each n N, let H n be a set in A such that H n H n+1. Show that P (H n ) monotonically converges to P ( k=1 H k) as n. (b) For each n

More information

Gaussian Processes. Le Song. Machine Learning II: Advanced Topics CSE 8803ML, Spring 2012

Gaussian Processes. Le Song. Machine Learning II: Advanced Topics CSE 8803ML, Spring 2012 Gaussian Processes Le Song Machine Learning II: Advanced Topics CSE 8803ML, Spring 01 Pictorial view of embedding distribution Transform the entire distribution to expected features Feature space Feature

More information

Analysis of Type-II Progressively Hybrid Censored Data

Analysis of Type-II Progressively Hybrid Censored Data Analysis of Type-II Progressively Hybrid Censored Data Debasis Kundu & Avijit Joarder Abstract The mixture of Type-I and Type-II censoring schemes, called the hybrid censoring scheme is quite common in

More information

Financial Econometrics

Financial Econometrics Financial Econometrics Estimation and Inference Gerald P. Dwyer Trinity College, Dublin January 2013 Who am I? Visiting Professor and BB&T Scholar at Clemson University Federal Reserve Bank of Atlanta

More information

University of California San Diego and Stanford University and

University of California San Diego and Stanford University and First International Workshop on Functional and Operatorial Statistics. Toulouse, June 19-21, 2008 K-sample Subsampling Dimitris N. olitis andjoseph.romano University of California San Diego and Stanford

More information

Testing Simple Hypotheses R.L. Wolpert Institute of Statistics and Decision Sciences Duke University, Box Durham, NC 27708, USA

Testing Simple Hypotheses R.L. Wolpert Institute of Statistics and Decision Sciences Duke University, Box Durham, NC 27708, USA Testing Simple Hypotheses R.L. Wolpert Institute of Statistics and Decision Sciences Duke University, Box 90251 Durham, NC 27708, USA Summary: Pre-experimental Frequentist error probabilities do not summarize

More information

Regression #5: Confidence Intervals and Hypothesis Testing (Part 1)

Regression #5: Confidence Intervals and Hypothesis Testing (Part 1) Regression #5: Confidence Intervals and Hypothesis Testing (Part 1) Econ 671 Purdue University Justin L. Tobias (Purdue) Regression #5 1 / 24 Introduction What is a confidence interval? To fix ideas, suppose

More information

Economics 583: Econometric Theory I A Primer on Asymptotics

Economics 583: Econometric Theory I A Primer on Asymptotics Economics 583: Econometric Theory I A Primer on Asymptotics Eric Zivot January 14, 2013 The two main concepts in asymptotic theory that we will use are Consistency Asymptotic Normality Intuition consistency:

More information

Lecture 3. Inference about multivariate normal distribution

Lecture 3. Inference about multivariate normal distribution Lecture 3. Inference about multivariate normal distribution 3.1 Point and Interval Estimation Let X 1,..., X n be i.i.d. N p (µ, Σ). We are interested in evaluation of the maximum likelihood estimates

More information

A Very Brief Summary of Statistical Inference, and Examples

A Very Brief Summary of Statistical Inference, and Examples A Very Brief Summary of Statistical Inference, and Examples Trinity Term 2008 Prof. Gesine Reinert 1 Data x = x 1, x 2,..., x n, realisations of random variables X 1, X 2,..., X n with distribution (model)

More information

A REVERSE TO THE JEFFREYS LINDLEY PARADOX

A REVERSE TO THE JEFFREYS LINDLEY PARADOX PROBABILITY AND MATHEMATICAL STATISTICS Vol. 38, Fasc. 1 (2018), pp. 243 247 doi:10.19195/0208-4147.38.1.13 A REVERSE TO THE JEFFREYS LINDLEY PARADOX BY WIEBE R. P E S T M A N (LEUVEN), FRANCIS T U E R

More information

F & B Approaches to a simple model

F & B Approaches to a simple model A6523 Signal Modeling, Statistical Inference and Data Mining in Astrophysics Spring 215 http://www.astro.cornell.edu/~cordes/a6523 Lecture 11 Applications: Model comparison Challenges in large-scale surveys

More information

Tests Concerning Equicorrelation Matrices with Grouped Normal Data

Tests Concerning Equicorrelation Matrices with Grouped Normal Data Tests Concerning Equicorrelation Matrices with Grouped Normal Data Paramjit S Gill Department of Mathematics and Statistics Okanagan University College Kelowna, BC, Canada, VV V7 pgill@oucbcca Sarath G

More information

Thomas J. Fisher. Research Statement. Preliminary Results

Thomas J. Fisher. Research Statement. Preliminary Results Thomas J. Fisher Research Statement Preliminary Results Many applications of modern statistics involve a large number of measurements and can be considered in a linear algebra framework. In many of these

More information

Evaluating the Performance of Estimators (Section 7.3)

Evaluating the Performance of Estimators (Section 7.3) Evaluating the Performance of Estimators (Section 7.3) Example: Suppose we observe X 1,..., X n iid N(θ, σ 2 0 ), with σ2 0 known, and wish to estimate θ. Two possible estimators are: ˆθ = X sample mean

More information