- Open Access
Admissibility in general linear model with respect to an inequality constraint under balanced loss
© Zhang and Gui; licensee Springer. 2014
- Received: 3 September 2013
- Accepted: 30 January 2014
- Published: 13 February 2014
Since Zellner (Bayesian and Non-Bayesian Estimation Using Balanced Loss Functions, pp. 377-390, 1994) proposed the balanced loss function, many researchers have been attracted to the field concerned. In this paper, under a generalized balanced loss function, we investigate the admissibility of linear estimators of the regression coefficient in general Gauss-Markov model (GGM) with respect to an inequality constraint. The necessary and sufficient conditions that the linear estimators of regression coefficient function are admissible are established, in the class of homogeneous/inhomogeneous linear estimation, respectively.
- general Gauss-Markov model
- homogeneous/inhomogeneous linear estimation
- inequality constraint
- balanced risk function
Throughout this paper, the symbols , , , , and stand for the transpose, the range, Moore-Penrose inverse, generalized inverse, rank, and trace of matrix A, respectively.
where y is a observable random vector. X is an known design matrix and , ε is a random error vector. β and are unknown parameters.
where , S is a known positive definite matrix. The balanced loss function takes both the precision of the estimator and the goodness-of-fit of the model into account. Compared to the standards in (1.2) and (1.3), it is a more comprehensive one that measures the estimate.
Much work has been done on the parameter estimation under the balanced loss function. [2–5] studied the risk function of some specific estimators. [6–8] did some work on the application of the balanced loss function. [9–13] investigated the goodness of the estimators under the balanced loss function.
where , S is a known matrix.
where r is a known vector. If , then the constraint condition always holds. This model embraces the unconstraint case.
and there exists , such that , where the risk function , then is said to be better than . If there does not exist any estimator in set Ξ that is better than , where parameters β and σ take values in T, then is called the admissible estimator of Kβ in the set Ξ. We denote it by .
where HL is the class of homogeneous linear estimators and L is the class of inhomogeneous linear estimators.
The admissibility is the most basic and influential rationality requirement of classical statistical decision theory. When the parameters are unconstrained, comprehensive results have been obtained. For instance, [14–17]etc. studied the admissibility in univariate linear model. As [18, 19] pointed out, when the parameters are constrained, the least square estimator may not be admissible. So it is significant to discuss the admissibility of linear estimator in linear model with some constraints. For the Gauss-Markov model with constraints,  developed the admissible estimator. Some other researchers dedicated to this study. [21–24] studied the admissibility in the linear model with an ellipsoidal constraint. For the linear model with an inequality constraint, [25–28] studied the admissibility of linear estimator of parameters in the univariate and multivariate linear models under the quadratic and matrix loss, respectively. However, under the balanced loss, the model with an inequality constraint has not been considered.
where is an estimator of .
where , .
where . Notice that , we have .
and the equality holds if and only if . □
Remark 2.1 This lemma indicates the class of estimators is a complete class of HL. That is, for any estimator δ not in , there exists an estimator in such that is better than δ.
Proof The lemma can easily be verified from (2.4). □
Lemma 2.4 Consider the model (1.5) with the loss function (1.6), suppose , then if and only if in model (2.6) with the loss function (2.7), where .
Equations (2.10), (2.11) and Lemma 2.3 indicate that if there exists an estimator of β, is better than , then , the estimator of Cβ, is better than . □
Lemma 2.5 Consider the model (2.6) with the loss function (2.7), holds if and only if under the quadratic loss, holds.
Proof The proof is straightforward. We omit the details. □
Theorem 2.1 Consider the model (2.6) with the loss function (2.7), if and only if .
Proof The necessity is trivial. We only need to prove the sufficiency. For any , if there exists that is better than AY, by Lemma 2.3, for any , (2.8) and (2.9) hold. Notice that (2.9) still holds if replacing β by −β. In other words, for any , (2.9) holds. Since , thus, for any , (2.8) and (2.9) hold. It contradicts with . □
Proof By Lemma 2.2, (1) holds. Further, is equivalent to . By Lemma 2.4, in the model (1.5) with the loss (1.6), holds if and only if in model (2.6) with the loss (2.7), where . It is also equivalent to in model (2.6) with the loss (2.1) by Lemma 2.5. Therefore, when the condition (1) is satisfied, according to Lemma 2.1 and simple computations, we have holds if and only if (2) and (3) are satisfied. □
Remark 2.2 The following example indicates that the conditions in the above theorem can be satisfied.
For the diagonal matrix , we consider the admissibility of Ay. The condition (1) in Theorem 2.2 is satisfied. Theorem 2.2(3) implies that . Theorem 2.2(2) implies that . Thus, only if and , Ay is an admissible estimate of β.
In this section, we study the admissibility in the class of inhomogeneous linear estimators.
if and only if and , where is the dual cone of C.
Proof This lemma can be found in . □
and the equality holds if and only if , . This means if , then is better than . It is a contradiction.
is better than , which contradicts .
It implies that no estimator is better than AY. Thus, . □
In fact, the converse part of Theorem 3.1 is also true. We present this in the following theorem.
Proof By the proof of (1) in Theorem 3.1, we need to prove that there does not exist matrix and such that is better than , where .
Plug (3.9), (3.10), and (3.14) into (3.7) and we find that the equality in (3.7) holds. It means there does not exist an estimator that is better than . Therefore, holds. □
We summarize Theorem 2.2 and Theorem 3.2 in the following theorem.
In this paper, under a generalized balanced loss function, we study the admissibility of linear estimators of the regression coefficient in general Gauss-Markov model with respect to an inequality constraint. The necessary and sufficient conditions that the linear estimators of regression coefficient function are admissible are obtained, in the class of homogeneous and inhomogeneous linear estimation, respectively.
This work was partially supported by National Natural Science Foundation of China (61070236, U1334211, 11371051) and the Project of State Key Laboratory of Rail Traffic Control and Safety (RCS2012ZT004), Beijing Jiaotong University.
- Zellner A: Bayesian and Non-Bayesian Estimation Using Balanced Loss Functions. Springer, Berlin; 1994:377–390.Google Scholar
- Rodrigues J, Zellner A: Weighted balanced loss function and estimation of the mean time to failure. Commun. Stat., Theory Methods 1994,23(12):3609–3616. 10.1080/03610929408831468MATHMathSciNetView ArticleGoogle Scholar
- Wan AT: Risk comparison of the inequality constrained least squares and other related estimators under balanced loss. Econ. Lett. 1994,46(3):203–210. 10.1016/0165-1765(94)00485-4View ArticleGoogle Scholar
- Giles JA, Giles DE, Ohtani K: The exact risks of some pre-test and stein-type regression estimators under balanced loss. Commun. Stat., Theory Methods 1996,25(12):2901–2924. 10.1080/03610929608831878MATHMathSciNetView ArticleGoogle Scholar
- Ohtani K: The exact risk of a weighted average estimator of the ols and stein-rule estimators in regression under balanced loss. Stat. Risk Model. 1998,16(1):35–46.MATHMathSciNetGoogle Scholar
- Shalabh : Least squares estimators in measurement error models under the balanced loss function. Test 2001,10(2):301–308. 10.1007/BF02595699MATHMathSciNetView ArticleGoogle Scholar
- Gruber DMH: The efficiency of shrinkage estimators with respect to Zellner’s balanced loss function. Commun. Stat., Theory Methods 2004,33(2):235–249. 10.1081/STA-120028372MATHView ArticleGoogle Scholar
- Akdeniz F, Wan AT, Akdeniz E: Generalized Liu type estimators under Zellner’s balanced loss function. Commun. Stat., Theory Methods 2005,34(8):1725–1736. 10.1081/STA-200066357MATHMathSciNetView ArticleGoogle Scholar
- Dey DK, Ghosh M, Strawderman WE: On estimation with balanced loss functions. Stat. Probab. Lett. 1999,45(2):97–101. 10.1016/S0167-7152(99)00047-4MATHMathSciNetView ArticleGoogle Scholar
- Ohtani K: Inadmissibility of the Stein-rule estimator under the balanced loss function. J. Econom. 1999,88(1):193–201. 10.1016/S0304-4076(98)00030-XMATHMathSciNetView ArticleGoogle Scholar
- Xu X, Wu Q: Linear admissible estimators of regression coefficient under balanced loss. Acta Math. Sci. 2000, 4: 468–473.Google Scholar
- Jozani JM, Marchand É, Parsian A: On estimation with weighted balanced-type loss function. Stat. Probab. Lett. 2006,76(8):773–780. 10.1016/j.spl.2005.10.026MATHView ArticleGoogle Scholar
- Cao M: ϕ admissibility for linear estimators on regression coefficients in a general multivariate linear model under balanced loss function. J. Stat. Plan. Inference 2009,139(9):3354–3360. 10.1016/j.jspi.2009.03.013MATHView ArticleGoogle Scholar
- Rao CR: Estimation of parameters in a linear model. Ann. Stat. 1976,4(6):1023–1037. 10.1214/aos/1176343639MATHView ArticleGoogle Scholar
- LaMotte LR: Admissibility in linear estimation. Ann. Stat. 1982,10(1):245–255. 10.1214/aos/1176345707MATHMathSciNetView ArticleGoogle Scholar
- Wu Q: Admissibility of linear estimators of regression coefficient in a general Gauss-Markoff model. Acta Math. Appl. Sin. 1986, 2: 251–256.Google Scholar
- Dong L, Wu Q: The sufficient and necessary conditions of admissible linear estimates for random regression coefficients and parameters under the quadratic loss function. Acta Math. Sin. 1988,31(2):145–157.MATHMathSciNetGoogle Scholar
- Marquaridt DW: Generalized inverses, ridge regression, biased linear estimation, and nonlinear estimation. Technometrics 1970,12(3):591–612.View ArticleGoogle Scholar
- Perlman MD: Reduced mean square error estimation for several parameters. Sankhyā, Ser. B 1972,34(1):89–92.MathSciNetGoogle Scholar
- Hoffmann K: Admissibility of linear estimators with respect to restricted parameter sets. Stat.: J. Theor. Appl. Stat. 1977,8(4):425–438.MATHGoogle Scholar
- Mathew T: Admissible linear estimation in singular linear models with respect to a restricted parameter set. Commun. Stat., Theory Methods 1985,14(2):491–498. 10.1080/03610928508828927MATHMathSciNetView ArticleGoogle Scholar
- Lu C: Admissibility of inhomogeneous linear estimators in linear models with respect to incomplete ellipsoidal restrictions. Commun. Stat., Theory Methods 1995,24(7):1737–1742. 10.1080/03610929508831582MATHView ArticleGoogle Scholar
- Zhang S, Gui W: Admissibility of linear estimators in a growth curve model subject to an incomplete ellipsoidal restriction. Acta Math. Sci. 2008,28(1):194–200. 10.1016/S0252-9602(08)60020-XMATHMathSciNetView ArticleGoogle Scholar
- Zhang S, Gui W, Liu G: Characterization of admissible linear estimators in the general growth curve model with respect to an incomplete ellipsoidal restriction. Linear Algebra Appl. 2009,431(1):120–131.MATHMathSciNetView ArticleGoogle Scholar
- Zhang S, Liu G, Gui W: Admissible estimators in the general multivariate linear model with respect to inequality restricted parameter set. J. Inequal. Appl. 2009., 2009: Article ID 718927Google Scholar
- Lu C, Shi N: Admissible linear estimators in linear models with respect to inequality constraints. Linear Algebra Appl. 2002,354(1):187–194.MATHMathSciNetView ArticleGoogle Scholar
- Zhang S, Fang Z, Qin H, Han L: Characterization of admissible linear estimators in the growth curve model with respect to inequality constraints. J. Korean Stat. Soc. 2011,40(2):173–179. 10.1016/j.jkss.2010.09.002MATHMathSciNetView ArticleGoogle Scholar
- Zhang S, Fang Z, Liu G: Characterization of admissible linear estimators in multivariate linear model with respect to inequality constraints under matrix loss function. Commun. Stat., Theory Methods 2013,42(15):2837–2850. 10.1080/03610926.2011.615441MATHMathSciNetView ArticleGoogle Scholar
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.