Skip to main content
Log in

Optimal shrinkage estimations in partially linear single-index models for binary longitudinal data

TEST Aims and scope Submit manuscript

Abstract

This paper focuses on the optimal estimation strategies of partially linear single-index models (PLSIM) for binary longitudinal data. Fitting model between the response and covariates may cause complexity and the linear terms may not be adequate to represent the relationship. In this situation, the PLSIM containing both linear and nonlinear terms is preferable. The objective of this paper is to develop optimal estimation strategies such as, pretest and shrinkage methods, for the analysis of binary longitudinal data under the PLSIM where some regression parameters are subject to restrictions. We estimate the nonparametric component using kernel estimating equations, and then use profile estimating equations to estimate the unrestricted and restricted estimators. To apply the pretest and shrinkage methods, we fit two models: one includes all covariates and the other restricts the regression parameters based on the auxiliary information. The unrestricted and restricted estimators are then combined optimally to get the pretest and shrinkage estimators. We also derive the asymptotic properties of the estimators in terms of biases and risks. Monte Carlo simulations are also conducted to examine the relative performance of the proposed estimators to the unrestricted estimator. An empirical application is also be used to illustrate the usefulness of our methodology.

This is a preview of subscription content, log in via an institution to check access.

Access this article

Price excludes VAT (USA)
Tax calculation will be finalised during checkout.

Instant access to the full article PDF.

Institutional subscriptions

Fig. 1
Fig. 2
Fig. 3

Similar content being viewed by others

References

  • Ahmed SE, Doksum K, Hossain S, You J (2007) Shrinkage, pretest and LASSO estimators in partially linear models. Aust N Z J Stat 49(4):461–471

    Article  Google Scholar 

  • Bai Y, Fung W, Zhu Z (2009) Penalized quadratic inference functions for single-index models with longitudinal data. J Multivar Anal 100:152–161

    Article  MathSciNet  Google Scholar 

  • Cai L, Wang S (2019) Inferences with generalized partially linear single-index models for longitudinal data. J Stat Plann Inference 200:146–160

    Article  MathSciNet  Google Scholar 

  • Cai Q, Wang S (2018) Efficient estimation in partially linear single-index models for longitudinal data. Scand J Stat 46:116–141

    Article  MathSciNet  Google Scholar 

  • Carroll RJ, Fan J, Gijbels I, Wand MP (1997) Genarlized partially linear single-index models. J Am Stat Assoc 92:477–489

    Article  Google Scholar 

  • Chen J, Gao J, Li D (2013) Estimation in partially linear single-index panel data models with fixed effects. J Bus Econom Stat 31:315–330

    Article  MathSciNet  Google Scholar 

  • Chen J, Li D, Liang H, Wang S (2015) Semiparametric GEE analysis in partially linear single-index models for longitudinal data. Ann Stat 43(4):1682–1715

    Article  MathSciNet  Google Scholar 

  • Chowdhury SK, Sinha SK (2015) Semiparametric marginal models for binary longitudinal data. Int J Stat Probab 4(3):107–120

    Article  Google Scholar 

  • Diggle P, Heagerty P, Liang KY, Zeger S (2013) Analysis of longitudinal data, 2nd edn. Oxford University Press, New York

    MATH  Google Scholar 

  • Fan J, Heckman NE, Wand MP (1995) Local polynomial kernel regression for generalized linear models and quasi-likelihood functions. J Am Stat Assoc 90:141–150

    Article  MathSciNet  Google Scholar 

  • Hoover DR, Rice JA, Wu CO, Yang Y (1998) Nonparametric smoothing estimates of time-varying coefficient models with longitudinal data. Biometrika 85:809–833

    Article  MathSciNet  Google Scholar 

  • Hossain S, Ahmed SE, Yi Y, Chen B (2016) Shrinkage and pretest estimators for longitudinal data analysis under partially linear models. J Nonparametr Stat 28:531–549

    Article  MathSciNet  Google Scholar 

  • Jiang CR, Wang JL (2011) Functional single index models for longitudinal data. Ann Stat 39:362–388

    Article  MathSciNet  Google Scholar 

  • Kadiri MA, Carroll RJ, Wand MP (2010) Marginal longitudinal semiparametric regression via penalized splines. Stat Probab Lett 80:1242–1252

    Article  MathSciNet  Google Scholar 

  • Lai P, Li G, Lian H (2013) Quadratic inference functions for partially linear single-index models with longitudinal data. J Multivar Anal 118:115–127

    Article  MathSciNet  Google Scholar 

  • Lawless JF, Singhal K (1978) Efficient screening of non-normal regression models. Biometrics 34:318–327

    Article  Google Scholar 

  • Li G, Zhu L, Xue L, Feng S (2010) Empirical likelihood inference in partially linear single-indexmodels for longitudinal data. J Multivar Anal 101:718–732

    Article  Google Scholar 

  • Li G, Lai P, Lian H (2015) Variable selection and estimation for partially linear single-index models with longitudinal data. Stat Comput 25:579–593

    Article  MathSciNet  Google Scholar 

  • Liang KY, Zeger SL (1986) Longitudinal data analysis using generalized linear models. Biometrika 73:13–22

    Article  MathSciNet  Google Scholar 

  • Lin H, Zhang R, Shi J, Liu J, Liu Y (2016) A new local estimation method for single index models for longitudinal data. J Nonparametr Stat 28:644–658

    Article  MathSciNet  Google Scholar 

  • Lin X, Carroll RJ (2001) Semiparametric regression for clustered data using generalized estimating equations. J Am Stat Assoc 96:1045–1056

    Article  MathSciNet  Google Scholar 

  • Lin X, Carroll RJ (2006) Semiparametric estimation in general repeated measures problems. J R Stat Soc Ser B Stat Methodol 68:68–88

    Article  MathSciNet  Google Scholar 

  • Moyeed RA, Diggle PJ (1994) Rates of convergence in semiparametric modelling of longitudinal data. Aust J Stat 36:75–93

    Article  Google Scholar 

  • Nkurunziza S, Chen F (2013) On extension of some identities for the bias and risk functions in elliptically contoured distributions. J Multivar Anal 122:190–201

    Article  MathSciNet  Google Scholar 

  • Ruppert D (1997) Empirical-bias bandwidths for local polynomial nonparametric regression and density estimation. J Am Stat Assoc 92:1049–1062

    Article  MathSciNet  Google Scholar 

  • Ruppert D, Sheater SJ, Wand MP (1995) An effective bandwidth selector for local least squares regression. J Am Stat Assoc 90:1257–1270

    Article  MathSciNet  Google Scholar 

  • Wang N, Carroll RJ, Lin X (2005) Efficient semiparametric marginal estimation for longitudinal/clustered data. J Am Stat Assoc 100(469):147–157

    Article  MathSciNet  Google Scholar 

  • Wu CO, Chiang CT (2000) Kernel smoothing on varying coefficient models with longitudinal dependent variable. Stat Sinica 10:433–456

    MathSciNet  MATH  Google Scholar 

  • Xu P, Zhu L (2012) Estimation for a marginal generalized single-index longitudinal model. J Multivar Anal 105:285–299

    Article  MathSciNet  Google Scholar 

  • Xu P, Zhang J, Huang X, Wang T (2016) Efficient estimation for marginal generalized partially linear single-index models with longitudinal data. TEST 25:413–431

    Article  MathSciNet  Google Scholar 

  • Yao W, Li R (2013) New local estimation procedure for a non-parametric regression function for longitudinal data. J R Stat Soc Ser B Stat Methodol 75:123–138

    Article  MathSciNet  Google Scholar 

  • Yi GY, Cook RJ (2002) Marginal methods for incomplete longitudinal data arising in clusters. J Am Stat Assoc 97:1071–1080

    Article  MathSciNet  Google Scholar 

  • Yi GY, He W, Liang H (2009) Analysis of correlated binary data under partially linear single-index logistic models. J Multivar Anal 100(2):278–290

    Article  MathSciNet  Google Scholar 

  • Zeger SL, Diggle PJ (1994) Semiparametric models for longitudinal data with application to CD4 cell numbers in HIV seroconverters. Biometrics 50:689–699

    Article  Google Scholar 

  • Zeger SL, Liang KY (1986) Longitudinal data analysis for discrete and continuous outcomes. Biometrics 42:121–130

    Article  Google Scholar 

Download references

Acknowledgements

The authors gratefully acknowledge an associate editor and two anonymous referees, whose comments greatly improved the form and content of this paper.The research of Dr. Shakhawat Hossain was supported by the Natural Sciences and Engineering Research Council of Canada.

Author information

Authors and Affiliations

Authors

Corresponding author

Correspondence to Shakhawat Hossain.

Additional information

Publisher's Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Appendix

Appendix

Proof of Theorem 4.2

Let \(\hat{\varvec{\xi }}=(\hat{\varvec{\xi }}_1^{\top }, \hat{\varvec{\xi }}_2^{\top })^{\top }\) be the solution of \(\sum _{i=1}^n S_i(\varvec{\xi }) = 0\). For parameter \(\varvec{\xi }_2\), we can show under \(H_0\) that \(\sqrt{n}(\hat{\varvec{\xi }}_2-\varvec{0})\) is multivariate normal with mean \(\varvec{0}\) and variance \({\varvec{{\mathcal {I}}}}_{\xi _{22}}\) as \(n\rightarrow \infty \). Hence \(n\hat{\varvec{\xi }}_2^{\top } {\varvec{{\mathcal {I}}}}_{\xi _{22}}^{-1} \hat{\varvec{\xi }}_2\) converges to \(\chi _{p_2}^2\) distribution.

Using the idea of the proof of Result 1 of Lin and Carroll (2001), it can be shown by Taylor expansion that \((\hat{\varvec{\xi }}-\varvec{\xi }) =n^{-1}\varvec{\varOmega }_{\varvec{\xi }} \varvec{S}(\varvec{\xi }_1, \varvec{\xi }_2) +o_p(1)\) and if \(H_0\) is true, we can write

$$\begin{aligned} \hat{\varvec{\xi }}_2 -\varvec{0} = \frac{1}{n} \left( \varvec{\varOmega }^{(1)} \varvec{S}_{\varvec{\xi }_1} (\varvec{\xi }_1,\varvec{0}) + \varvec{\varOmega }_{\varvec{\xi }_{22}} \varvec{S}_{\varvec{\xi }_2} (\varvec{\xi }_1,\varvec{0})\right) +o_p(1) , \end{aligned}$$

where \(\varvec{\varOmega }^{(1)}\) is the lower left of \(p_2\times (q+p_1)\) submatrix of \(\varvec{\varOmega }_{\varvec{\xi }}\). Plug-in the value of \(\hat{\varvec{\xi }}_2\) in the following expression

$$\begin{aligned}&n\hat{\varvec{\xi }}_{2}^{\top } {\varvec{{\mathcal {I}}}}_{{\varvec{\xi }}_{22}}^{-1}\hat{\varvec{\xi }}_{2} \nonumber \\&\quad = n^{-1} {\varvec{S}}_{ {\varvec{\xi }}_1}^{\top }(\varvec{\xi }_1, \varvec{0}) \left( [{\varvec{\varOmega }}^{(1)}]^{\top } {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22}}^{-1} {\varvec{\varOmega }}^{(1)}\right) {\varvec{S}}_{\varvec{\xi }_1}( \varvec{\xi }_1, \varvec{0}) \nonumber \\&\qquad + n^{-1} {\varvec{S}}_{\varvec{\xi }_2}^{\top }(\varvec{\xi }_1, \varvec{0}) {\varvec{\varOmega }}_{\varvec{\xi }_{22}}^{\top } {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22}}^{-1} {\varvec{\varOmega }}^{(1)} {\varvec{S}}_{\varvec{\xi }_1}(\varvec{\xi }_1, \varvec{0}) \nonumber \\&\qquad +n^{-1} {\varvec{S}}_{\varvec{\xi }_1}^{\top }(\varvec{\xi }_1, \varvec{0}) \left( [{\varvec{\varOmega }}^{(1)}]^{\top } {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22}}^{-1} {\varvec{\varOmega }}_{\varvec{\xi }_{22}}\right) {\varvec{S}}_{\varvec{\xi }_2}(\varvec{\xi }_1, \varvec{0}) \nonumber \\&\qquad + n^{-1} {\varvec{S}}_{\varvec{\xi }_2}^{\top }(\varvec{\xi }_1, \varvec{0})\left( {\varvec{\varOmega }}_{\varvec{\xi }_{22}}^{\top } {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22}}^{-1} {\varvec{\varOmega }}_{\varvec{\xi }_{22}}\right) {\varvec{S}}_{\varvec{\xi }_2}( \varvec{\xi }_1, \varvec{0}). \end{aligned}$$
(9)

which has an asymptotic \(\chi _{p_2}^2\) distribution with \(p_2\) degrees of freedom. To show this, assume that \({\hat{\varXi }}_L^*\) be the new version of relation (9) when \( \varvec{\xi }_1\) is replaced by \(\tilde{\varvec{\xi }}_1\). Now the relation (9) differs from \({\hat{\varXi }}_L^*\) by an \(o_p(1)\) term as \(n\rightarrow \infty \). Thus \({\hat{\varXi }}_L^*\) converges in distribution to \(\chi _{p_2}^2\). We know that \({\varvec{S}}_{\varvec{\xi }_1}(\tilde{\varvec{\xi }}_1, \varvec{0})=\varvec{0}\) and this \({\varvec{S}}_{\varvec{\xi }_1}(\tilde{\varvec{\xi }}_1, \varvec{0})\) involves in the first three terms of (9) when \( \varvec{\xi }_1\) is replaced by \(\tilde{\varvec{\xi }}_1\). Thus \({\hat{\varXi }}_L^*= {\hat{\varXi }}_L\) and hence we complete the proof of the theorem.

Under local alternative (6), the following theorem and Lemma facilitates the theoretical derivation of ADB and ADR of the RPSIM, PTSIM, SESIM, and PSESIM estimators in Theorems 4.1.2 and 4.1.3. \(\square \)

Proof of Theorem 4.1.1

$$\begin{aligned} \text{ E }(\varvec{\psi }_1)= & {} \text{ E }\left( \sqrt{n}(\hat{\varvec{\xi }} - \varvec{\xi })\right) = \varvec{0}.\\ \text{ E }(\varvec{\psi }_2)= & {} \text{ E }\left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi })\right) \\= & {} \text{ E }\left( \sqrt{n}(\hat{\varvec{\xi }}_1- \varvec{\xi }_1+ {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} \hat{\varvec{\xi }}_2)\right) ,~~\text{ see } Lawless and Singhal (1978)\\= & {} \varvec{0} + {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} \sqrt{n}\left( \frac{\varvec{\delta }}{\sqrt{n}}\right) ={\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}}\varvec{\delta } =\varvec{\gamma }. \end{aligned}$$
$$\begin{aligned} \text{ E }(\varvec{\psi }_3)= & {} \text{ E }\left( \sqrt{n}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})\right) = \text{ E }\left( \sqrt{n}(\hat{\varvec{\xi }} - \varvec{\xi })-\sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi })\right) = - \varvec{\gamma }.\\ \text{ Var }(\varvec{\psi }_1)= & {} \text{ Var }\left( \sqrt{n}(\hat{\varvec{\xi }} - \varvec{\xi })\right) = {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11.2}}^{-1} ={\mathcal {A}}_{11}.\\ \text{ Var }(\varvec{\psi }_2)= & {} \text{ Var }\left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi })\right) =\text{ Var }\left( \sqrt{n}(\hat{\varvec{\xi }}_1- \varvec{\xi }_1+ {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} \hat{\varvec{\xi }}_2)\right) \\= & {} \text{ Var }\left( \sqrt{n}\left( \hat{\varvec{\xi }}_1- \varvec{\xi }_1\right) \right) + {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} \text{ Var }\left( \sqrt{n} \hat{\varvec{\xi }}_2\right) {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} \\&+ 2 \text{ Cov }\left( \sqrt{n}\left( \hat{\varvec{\xi }}_1- \varvec{\xi }_1\right) , {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} \sqrt{n} \hat{\varvec{\xi }}_2\right) \\= & {} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11.2}}^{-1} - {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22.1}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} = {\mathcal {A}}_{22}. \end{aligned}$$
$$\begin{aligned} \text{ Var }(\varvec{\psi }_3)= & {} \text{ Var }\left( \sqrt{n}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})\right) = \text{ Var }\left( \sqrt{n}(\hat{\varvec{\xi }} - \varvec{\xi })-\sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi })\right) \\= & {} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22.1}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} = {\mathcal {A}}_{33}.\\ \text{ Cov }(\varvec{\psi }_1, \varvec{\psi }_2)= & {} \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi })\right) \\= & {} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11.2}}^{-1} - {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22.1}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} = {\mathcal {A}}_{12} = {\mathcal {A}}_{21}^{\top }.\\ \text{ Cov }(\varvec{\psi }_1, \varvec{\psi }_3)= & {} \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})\right) \\= & {} \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi })-\sqrt{n} (\tilde{\varvec{\xi }} - \varvec{\xi })\right) \\= & {} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22.1}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1}= {\mathcal {A}}_{13} = {\mathcal {A}}_{31}^{\top } \end{aligned}$$
$$\begin{aligned} \text{ Cov }(\varvec{\psi }_2, \varvec{\psi }_3)= & {} \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})\right) \\= & {} \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\hat{\varvec{\xi }}-\varvec{\xi }) - \sqrt{n}(\tilde{\varvec{\xi }}-\varvec{\xi })\right) \\= & {} \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\hat{\varvec{\xi }}-\varvec{\xi })\right) - \text{ Cov } \left( \sqrt{n}(\tilde{\varvec{\xi }} - \varvec{\xi }), \sqrt{n}(\tilde{\varvec{\xi }}-\varvec{\xi })\right) \\= & {} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11.2}}^{-1} - {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22.1}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} -{\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11.2}}^{-1} \\&\qquad + {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{22.1}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{21}} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1}=\varvec{0}. \end{aligned}$$

\(\square \)

Lemma

Let \(\varvec{X}\sim \text{ n } (\varvec{\mu }, \varvec{{\mathcal {I}}})\), where \(\varvec{{\mathcal {I}}}\) is a nonnegative definite matrix. Also let \(\varvec{Q}\) be symmetric and positive definite matrix such that \(\varvec{Q}^{1/2} \varvec{{\mathcal {I}}} \varvec{Q}^{1/2}\) is an idempotent matrix, and \(\varvec{Q} \varvec{{\mathcal {I}}} \varvec{Q}\varvec{\mu }=\varvec{Q}\varvec{\mu }\). Then for all \(\varphi \), Borel measurable and real-valued integrable function

$$\begin{aligned}&1 \text{ E }\left( \varphi \left( \varvec{X}^{\top } \varvec{Q} \varvec{{\mathcal {I}}} \varvec{Q}\varvec{X}\right) \varvec{X}\right) = \varvec{\theta }\text{ E }\left( \varphi \left( \chi _{p+2}^2(\varvec{\mu }^{\top } \varvec{Q} \varvec{\mu }) \right) \right) \\&2 \text{ E }\left( \varphi \left( \varvec{X}^{\top } \varvec{Q} \varvec{{\mathcal {I}}} \varvec{Q}\varvec{X}\right) \varvec{X}^{\top }\varvec{M}\varvec{X}\right) = \text{ E }\left( \varphi \left( \chi _{p+2}^2(\varvec{\mu }^{\top } \varvec{Q} \varvec{\mu }) \right) \right) \text{ tr }(\varvec{A\varvec{{\mathcal {I}}}})\\&\qquad +\text{ E }\left( \varphi \left( \chi _{p+4}^2(\varvec{\mu }^{\top } \varvec{Q} \varvec{\mu }) \right) \right) \varvec{\mu }^{\top } \varvec{M} \varvec{\mu }, \end{aligned}$$

where \(\varvec{M}\) is nonnegative definite matrix.

The outline of the proof of this lemma is given in Nkurunziza et al. (2013).

Proof of Theorem 4.1.2

In this proof, we derive the bias expressions of the proposed estimators. It is obvious that \(\text{ ADB }(\hat{\varvec{\xi }})=\varvec{0} \). The ADB of RPSIM (\(\tilde{\varvec{\xi }}\)), PTSIM (\(\hat{\varvec{\xi }}_p\)), SESIM(\(\hat{\varvec{\xi }}_S\)), and PSESIM (\(\hat{\varvec{\xi }}_{S+}\)) estimators are as follows:

$$\begin{aligned} \text{ ADB }(\tilde{\varvec{\xi }})&= \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\tilde{\varvec{\xi }} - \varvec{\xi }) \right) = \text{ E }(\varvec{\psi }_2) = -{\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{11}}^{-1} {\varvec{{\mathcal {I}}}}_{\varvec{\xi }_{12}} \varvec{\delta } = \varvec{\gamma }\\ \text{ ADB }(\hat{\varvec{\xi }}_{P})&= \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\hat{\varvec{\xi }}_{P} - \varvec{\xi }) \right) = \varvec{0}- \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} I\left( {\hat{\varXi }}_L \le \chi ^2_{\kappa +2, \alpha }\right) \right) \left( \hat{\varvec{\xi }} - \tilde{\varvec{\xi }}\right) \\&= - \text{ E } \left( \lim _{n\rightarrow \infty } I\left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }, \varDelta \right) \varvec{\psi }_3 \right) = \varvec{\gamma } H_{\kappa +4} \left( \chi ^2_{\kappa +2, \alpha }, \varDelta \right) \\ \text{ ADB }(\hat{\varvec{\xi }}_{S})&= \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\hat{\varvec{\xi }}_{S} - \varvec{\xi }) \right) = - \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\kappa {\hat{\varXi }}_L^{-1}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})) \right) \\&= -\kappa \text{ E } \left( \lim _{n\rightarrow \infty } \varvec{\psi }_3 {\hat{\varXi }}_L^{-1}\right) =\kappa \varvec{\gamma } \text{ E }(Z_1^{-1})\\ \text{ ADB }(\hat{\varvec{\xi }}_{S+})&= \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\hat{\varvec{\xi }}_{S+} - \varvec{\xi }) \right) \\&= \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n}(\hat{\varvec{\xi }}_{S} - \varvec{\xi }) - \sqrt{n} \left( 1-\kappa {\hat{\varXi }}_L^{-1}\right) I\left( {\hat{\varXi }}_L< \kappa \right) (\hat{\varvec{\xi }}-\tilde{\varvec{\xi }}) \right) \\&= \text{ ADB }(\hat{\varvec{\xi }}_{S}) - \text{ E } \left( \lim _{n\rightarrow \infty } \varvec{\psi }_3 \left( 1-\kappa {\hat{\varXi }}_L^{-1}\right) I\left( {\hat{\varXi }}_L< \kappa \right) \right) \\&= \text{ ADB }(\hat{\varvec{\xi }}_{S})+ \varvec{\gamma } \text{ E } \left( I\left( {\hat{\varXi }}_L< \kappa \right) \right) - \varvec{\gamma }\kappa \text{ E }\left( \varvec{\psi }_3 {\hat{\varXi }}_L^{-1}I\left( {\hat{\varXi }}_L< \kappa \right) \right) \\&= \text{ ADB }(\hat{\varvec{\xi }}_{S})+ \varvec{\gamma } H_{\kappa +4}(\kappa ,\varDelta ) - \varvec{\gamma }\kappa \text{ E }\left( Z_1^{-1} I\left( Z_1< \kappa \right) \right) \end{aligned}$$

Proof of Theorem 4.1.3

Based on the definition of ADR function, it is necessary to derive the asymptotic covariance matrices for the four estimators. The covariance matrix of any estimator \(\hat{\varvec{\xi }}^*\) is defined as:

$$\begin{aligned} \text{ Cov } (\hat{\varvec{\xi }}^*) = \text{ E }\left( \lim _{n\rightarrow \infty } n (\hat{\varvec{\xi }}^*-\varvec{\xi }) (\hat{\varvec{\xi }}^*-\varvec{\xi })^{\top } \right) . \end{aligned}$$

First, we will start deriving the covariance matrices of the UPSIM and RPSIM:

$$\begin{aligned} \text{ Cov } (\hat{\varvec{\xi }})= & {} \text{ E }\left( \lim _{n\rightarrow \infty } \sqrt{n} (\hat{\varvec{\xi }} -\varvec{\xi }) \sqrt{n}( \hat{\varvec{\xi }} -\varvec{\xi })^{\top } \right) = \text{ E }(\varvec{\psi } \varvec{\psi } ^{\top }) = {\varvec{{\mathcal {A}}}}_{12} ={\varvec{{\mathcal {I}}}}^{-1}_{11.2}.\\ \text{ Cov }( \tilde{\varvec{\xi }})= & {} \text{ E }\left( \lim _{n\rightarrow \infty } \sqrt{n}(\tilde{\varvec{\xi }} -\varvec{\xi }) \sqrt{n}(\tilde{\varvec{\xi }} -\varvec{\xi })^{\top }\right) = \text{ E }(\varvec{\psi }_2 \varvec{\psi }_2^{\top })= {\varvec{{\mathcal {A}}}}_{12} +\varvec{\gamma } \varvec{\gamma }^{\top }. \end{aligned}$$

Second, we derive the covariance matrix of the pretest estimator:

$$\begin{aligned} \text{ Cov }( \hat{\varvec{\xi }}_{P})= & {} \text{ E }\left( \lim _{n\rightarrow \infty } \sqrt{n}(\hat{\varvec{\xi }}_{P} - \varvec{\xi }) \sqrt{n} (\hat{\varvec{\xi }}_{P} - \varvec{\xi })^{\top } \right) \\= & {} \text{ E } \left( \varvec{\psi }_1 \varvec{\psi }_1^{\top } + \varvec{\psi }_3 \varvec{\psi }_3^{\top } \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) -2 \varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) \\= & {} {\varvec{{\mathcal {A}}}_{11}} + {\varvec{{\mathcal {A}}}_{13}} H_{\kappa + 4}\left( \chi ^2_{\kappa + 2,\alpha }, \varDelta \right) + \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 6}\left( \chi ^2_{\kappa + 2,\alpha }, \varDelta \right) \\&- 2 \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } I\left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) . \end{aligned}$$

Consider the fourth term:

$$\begin{aligned}&\text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) =\text{ E } \left( \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) | \varvec{\psi }_3\right) \right) \\&\quad = \text{ E } \left( \varvec{\psi }_3 \left( \varvec{\psi }_3^{\top } - \text{ E }\left( \varvec{\psi }_3\right) ^{\top }\right) \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) \\&\quad = \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_3^{\top } \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) -\text{ E } \left( \varvec{\psi }_3 \lim _{n\rightarrow \infty } I \left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) \text{ E }\left( \varvec{\psi }_3\right) ^{\top } \\&\quad = {\varvec{{\mathcal {A}}}_{13}} H_{\kappa +4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) + \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa +6}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) \\&\qquad - \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa +4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) \\&\quad = {\varvec{{\mathcal {A}}}_{13}} H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) + \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 6}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) - \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) ,\\ \text{ Cov }( \hat{\varvec{\xi }}_{P})= & {} {\varvec{{\mathcal {I}}}}^{-1}_{11.2}+ \lim _{n\rightarrow \infty } \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_3^{\top } I\left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) -2 \lim _{n\rightarrow \infty } \text{ E } \left( \varvec{\psi }_3 \varvec{\psi } ^{\top } I\left( {\hat{\varXi }}_L \le \chi ^2_{p_2, \alpha }\right) \right) \\&\quad = {\varvec{{\mathcal {I}}}}^{-1}_{11.2} + {\varvec{{\mathcal {A}}}_{13}} H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) + \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 6}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) \\&\qquad -2 \left( {\varvec{{\mathcal {A}}}_{13}} H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) + \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 6}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) \right. \\&\left. -\varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) \right) \\&\quad = {\varvec{{\mathcal {I}}}}^{-1}_{11.2} - {\varvec{{\mathcal {A}}}_{13}} H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) - \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 6}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) \\&\qquad + 2 \varvec{\gamma } \varvec{\gamma }^{\top } H_{\kappa + 4}\left( \chi ^2_{\kappa + 2, \alpha }, \varDelta \right) . \end{aligned}$$

Third, we derive the covariance matrices of the shrinkage and positive shrinkage estimators:

$$\begin{aligned} \text{ Cov } ( \hat{\varvec{\xi }}_{S})= & {} \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n}(\hat{\varvec{\xi }}_{S} - \varvec{\xi }) \sqrt{n} (\hat{\varvec{\xi }}_{S} - \varvec{\xi })^{\top } \right) \\= & {} \text{ E } \Bigg (\lim _{n\rightarrow \infty } \sqrt{n} \left( \hat{\varvec{\xi }} - \varvec{\xi } - \kappa {\hat{\varXi }}_L^{-1}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})\right) \sqrt{n} \left( \hat{\varvec{\xi }} - \varvec{\xi } - \kappa {\hat{\varXi }}_L^{-1}(\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})\right) ^{\top } \Bigg ) \\= & {} \text{ E } (\varvec{\psi }_1 \varvec{\psi }_1^{\top }) + \kappa ^2 \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_3^{\top } \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-2}\right) -2 \kappa \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1}\right) \\= & {} {\varvec{{\mathcal {I}}}}^{-1}_{11.2} + {\varvec{{\mathcal {A}}}_{13}} \text{ E }\left( Z_1^{-2}\right) + \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E }\left( Z_2^{-2}\right) -2 \kappa \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1}\right) . \end{aligned}$$

Consider the last term:

$$\begin{aligned}&\text{ E } (\varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1})=\text{ E } \left( \text{ E } (\varvec{\psi }_3 \varvec{\psi }_1^{\top } \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1} | \varvec{\psi }_3)\right) \\&\quad = \text{ E } \left( \varvec{\psi }_3 \text{ E } (\varvec{\psi }_1^{\top } | \varvec{\psi }_3) \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1}\right) + \text{ E } \left( \varvec{\psi }_3 \left( \varvec{\psi }_3^{\top } - \text{ E }(\varvec{\psi }_3\right) ^{\top }) \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1}\right) \\&\quad = \text{ E } \left( \varvec{\psi }_3 \varvec{\psi }_3^{\top } \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1}\right) - \text{ E } \left( \varvec{\psi }_3 \lim _{n\rightarrow \infty } {\hat{\varXi }}_L^{-1}\right) \text{ E }\left( \varvec{\psi }_3\right) ^{\top } \\&\quad = {\varvec{{\mathcal {A}}}_{13}} \text{ E } \left( \chi ^{-2}_{p_2 + 2}(\varDelta )\right) + \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } \left( \chi ^{-2}_{p_2 + 4}(\varDelta )\right) - \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } \left( \chi ^{-2}_{p_2 + 2}(\varDelta )\right) \\&\quad = {\varvec{{\mathcal {A}}}_{13}} \text{ E } \left( Z_1^{-1}\right) + \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } \left( Z_2^{-1}\right) - \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } \left( Z_1^{-1}\right) . \end{aligned}$$

Hence

$$\begin{aligned} \text{ Cov } ( \hat{\varvec{\xi }}_{S})= & {} {\varvec{{\mathcal {I}}}}^{-1}_{11.2} + \kappa ^2 \left( {\varvec{{\mathcal {A}}}_{13}} \text{ E } (Z_1^{-2}) + \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E }(Z_2^{-2}) \right) \\&-2 \kappa \left( {\varvec{{\mathcal {A}}}_{13}} \text{ E }(Z_1^{-1}) + \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } (Z_2^{-1}) - \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } (Z_1^{-1})\right) \\= & {} {\varvec{{\mathcal {I}}}}^{-1}_{11.2} + \left( \kappa ^2 \text{ E } \left( Z_1^{-2}\right) -2 \kappa \text{ E } (Z_1^{-1}) \right) {\varvec{{\mathcal {A}}}_{13}}\\&+ \left( \kappa ^2 \text{ E } (Z_2^{-2}) +2 \kappa \text{ E } (Z_1^{-1}) -2 \kappa \text{ E } (Z_2^{-1}) \right) \varvec{\gamma } \varvec{\gamma }^{\top }. \end{aligned}$$

Let \(F_m(\varDelta )=\left( 1-\kappa {\hat{\varXi }}_L^{-1}\right) ^m I \left( {\hat{\varXi }}_L < \kappa \right) \), where \(m=1,2\)

$$\begin{aligned} \text{ Cov } (\hat{\varvec{\xi }}_{S+})= & {} \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\hat{\varvec{\xi }}_{S+} - \varvec{\xi }) \sqrt{n} (\hat{\varvec{\xi }}_{S+} - \varvec{\xi })^{\top }\right) , \\= & {} \text{ E } \left( \lim _{n\rightarrow \infty } \sqrt{n} (\hat{\varvec{\xi }}_{S} - \varvec{\xi }) \sqrt{n} (\hat{\varvec{\xi }}_{S} - \varvec{\xi })^{\top }\right) \\&+ \text{ E } \left( \lim _{n\rightarrow \infty } F_2(\varDelta ) \sqrt{n} (\hat{\varvec{\xi }} - \tilde{\varvec{\xi }}) \sqrt{n} (\hat{\varvec{\xi }} - \tilde{\varvec{\xi }})^{\top }\right) \\&-2 \text{ E } \left( \lim _{n\rightarrow \infty } F_1(\varDelta ) \sqrt{n} (\hat{\varvec{\xi }} - \tilde{\varvec{\xi }}) \sqrt{n}(\hat{\varvec{\xi }}_{S} - \varvec{\xi })^{\top }\right) \\= & {} \text{ Cov } ( \hat{\varvec{\xi }}_{S}) + \text{ E } \left( \lim _{n\rightarrow \infty } F_2(\varDelta ) \varvec{\psi }_3 \varvec{\psi }_2^{\top }\right) \\- & {} 2 \text{ E } \left( \lim _{n\rightarrow \infty } F_1(\varDelta ) \varvec{\psi }_3 \left( \varvec{\psi }_2^{\top } + \left( 1-\kappa {\hat{\varXi }}_L^{-1}\right) \varvec{\psi }_3^{\top }\right) \right) , \\= & {} \text{ Cov } ( \hat{\varvec{\xi }}_{S}) - \text{ E } \left( \lim _{n\rightarrow \infty }F_2(\varDelta )\varvec{\psi }_3 \varvec{\psi }_3^{\top }\right) -2 \text{ E } \left( \lim _{n\rightarrow \infty } F_1(\varDelta ) \varvec{\psi }_3 \varvec{\psi }_2^{\top }\right) . \end{aligned}$$

\(\square \)

Consider the second term:

$$\begin{aligned}&- \text{ E } \left( \lim _{n\rightarrow \infty } F_2(\varDelta )\varvec{\psi }_3 \varvec{\psi }_3^{\top }\right) = - \text{ E } \left( \lim _{n\rightarrow \infty } \left( 1-\kappa {\hat{\varXi }}_L^{-1}\right) ^2 I \left( {\hat{\varXi }}_L< \kappa \right) \varvec{\psi }_3 \varvec{\psi }_3^{\top }\right) \\&\quad = - {\varvec{{\mathcal {A}}}_{13}} \text{ E } \left( I \left( Z_1< \kappa \right) \left( 1 - \kappa Z_1^{-1} \right) ^2\right) - \varvec{\gamma } \varvec{\gamma }^{\top } \text{ E } \left( I \left( Z_2 < \kappa \right) \left( 1 - \kappa Z_2^{-1} \right) ^2\right) . \end{aligned}$$

Consider the third term:

$$\begin{aligned}&-2 \text{ E } \left( \lim _{n\rightarrow \infty } F_1(\varDelta ) \varvec{\psi }_3 \varvec{\psi }_2^{\top }\right) = -2 \text{ E } \left( \lim _{n\rightarrow \infty } \varvec{\psi }_3 \text{ E } \left( F_1(\varDelta ) \varvec{\psi }_2^{\top } | \varvec{\psi }_3 \right) \right) \\&\quad = -2 \text{ E } \left( \lim _{n\rightarrow \infty } \varvec{\psi }_3 \left( \text{ E } \left( \varvec{\psi }_2^{\top }\right) + \text{ cov } \left( \varvec{\psi }_3, \varvec{\psi }_2\right) \left( \varvec{\psi }_3 - \text{ E }\left( \varvec{\psi }_3\right) \right) \right) F_1(\varDelta )\right) \\&\quad = -2 \text{ E } \left( \lim _{n\rightarrow \infty } \varvec{\psi }_3 \text{ E } \left( \varvec{\psi }_2^{\top }\right) F_1(\varDelta ) + \varvec{0}\right) \\&\quad = -2 \text{ E } \left( \lim _{n \rightarrow \infty } \varvec{\psi }_3 I \left( {\hat{\varXi }}_L< \kappa \right) - \kappa {\hat{\varXi }}_L^{-1} \varvec{\psi }_3 I \left( {\hat{\varXi }}_L< \kappa \right) \right) \text{ E } \left( \varvec{\psi }_2^{\top }\right) \\&\quad = 2 H_{\kappa + 4} (\kappa , \varDelta ) \varvec{\gamma } \varvec{\gamma }^{\top } - 2 \kappa \text{ E } \left( Z_1^{-1} I \left( Z_1< \kappa \right) \right) \varvec{\gamma } \varvec{\gamma }^{\top } \\&\quad = \left( 2 H_{\kappa + 4} (\kappa , \varDelta ) - 2 \kappa \text{ E } \left( Z_1^{-1} I \left( Z_1< \kappa \right) \right) \right) \varvec{\gamma } \varvec{\gamma }^{\top }. \end{aligned}$$

Finally,

$$\begin{aligned} \text{ Cov } \left( \hat{\varvec{\xi }}_{S+}\right)= & {} \text{ Cov } \left( \hat{\varvec{\xi }}_{S}\right) - \text{ E } \left( \left( 1 - \kappa Z_1^{-1}\right) ^2 I \left( Z_1< \kappa \right) \right) {\varvec{{\mathcal {A}}}_{13}} \\&+ \left( 2 H_{\kappa + 4} (\kappa , \varDelta ) - 2 \kappa \text{ E } \left( Z_1^{-1} I \left( Z_1< \kappa \right) \right) \right. \\&- \left. \text{ E } \left( \left( 1 - \kappa Z_2^{-1} \right) ^2 I \left( Z_2 < \kappa \right) \right) \right) \varvec{\gamma } \varvec{\gamma }^{\top }. \end{aligned}$$

The ADR expressions in Theorem 4.4 now follow from (8) which completes the proof.

Rights and permissions

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

Hossain, S., Lac, L.A. Optimal shrinkage estimations in partially linear single-index models for binary longitudinal data. TEST 30, 811–835 (2021). https://doi.org/10.1007/s11749-021-00753-3

Download citation

  • Received:

  • Accepted:

  • Published:

  • Issue Date:

  • DOI: https://doi.org/10.1007/s11749-021-00753-3

Keywords

Mathematics Subject Classification

Navigation