Abstract
Panel count data refer to interval-censored recurrent event data. Each study subject can only be observed at discrete time points, leading to knowledge about the total number of events occurring between observations. The observation times can be also different among subjects and carry important information about the underlying recurrent process. In this paper, an empirical likelihood (EL) method for panel count data with informative observation times is proposed. Based on the influence function, we formulate an empirical likelihood ratio for the vector of regression coefficients, and the Wilks’ theorem is established. Simulation studies are carried out to compare the performance of empirical likelihood with normal approximation methods. Finally, the EL method is compared with existing approaches, utilizing an illustrative example drawn from a bladder cancer study.
Similar content being viewed by others
References
Andersen PK, Gill RD (1982) Cox’s regression model for counting processes: a large sample study. Ann Stat 10(4):1100–1120
Andersen PK, Borgan O, Gill RD, Keiding N (1993) Statistical models based on counting processes. Springer-Verlag, New York
Andrews DF, Herzberg AM (1985) Data: a collection of problems from many fields for the student and research worker. Springer-Verlag, New York
Balakrishnan N, Zhao X (2009) New multi-sample nonparametric tests for panel count data. Ann Stat 37(3):1112–1149
Buzkova P (2010) Panel count data regression with informative observation times. Int J Biostat 6(1):30
Chen SX, Cui HJ (2006) On Bartlett correction of empirical likelihood in the presence of nuisance parameters. Bimetrika 16:1101–1115
Cook RJ, Lawless JF (2007) The statistical analysis of recurrent events. Springer-Verlag, New York
Dauxois J-V, Flesch A, Varron D (2016) Empirical likelihood confidence bands for mean functions of recurrent events with competing risks and a terminal event. ESAIM 20:66–94
DiCiccio TJ, Hall P, Romano JP (1991) Empirical likelihood is Bartlett-corrrectable. Ann Stat 19:1053–1061
Fang S, Zhang H, Sun L, Wang D (2017) Analysis of panel count data with time-dependent covariates and informative observation process. Acta Math Appl Sin 33(1):147–156
Fleming TR, Harrington DP (1991) Counting processes and survival analysis. In: Wiley series in probability and mathematical statistics: applied probability and statistics. Wiley, New York
Hall P, Scala BL (1990) Methodology and algorithms of empirical likelihood. Int Stat Rev 58:109–127
He S, Liang W, Shen J, Yang G (2016) Empirical likelihood for right censored lifetime data. J Am Stat Assoc 111:646–655
He X, Tong X, Sun J, Cook R (2008) Regression analysis of multivariate panel count data. Biostatistics 9:234–248
He X, Tong X, Sun J (2009) Semiparametric analysis of panel count data with correlated observation and follow-up times. Lifetime Data Anal 15:177–196
Hu S, Lin L (2012) Empirical likelihood analysis of longitudinal data involving within-subject correlation. Acta Math Appl Sin Engl Ser 28(4):731–744
Huang CY, Wang MC, Zhang Y (2006) Analysing panel count data with informative observation times. Biometrika 93(4):763–775
Kalbfleisch JD, Lawless JF (1985) The analysis of panel data under a Markov assumption. J Am Stat Assoc 80:863–871
Li N, Sun L, Sun J (2010) Semiparametric transformation models for panel count data with dependent observation process. Stat Biosci 2(2):191–210
Lin DY, Wei LJ, Yang I, Ying Z (2000) Semiparametric regression for the mean and rate functions of recurrent events. J R Stat Soc B 62(4):711–730
Liu P, Zhao Y (2023) A review of recent advances in empirical likelihood. WIREs Comput Stat 15(3):e1599
Liu W, Lu X, Xie C (2014) Empirical likelihood for the additive hazards model with current status data. Commun Stat 45(8):2720–2732
Owen AB (1988) Empirical likelihood ratio confidence intervals for a single functional. Biometrika 75:237–249
Owen AB (1990) Empirical likelihood ratio confidence regions. Ann Stat 18:90–120
Owen AB (1991) Empirical likelihood for linear models. Ann Stat 19:1725–1747
Owen AB (2001) Empirical likelihood. Chapman and Hall-CRC, New York
Qin J, Lawless J (1994) Empirical likelihood and general estimating equations. Ann Stat 22:300–325
Satter F (2020) Novel empirical likelihood inference procedures for zero-inflated and right censored data and their applications. PhD Dissertation, Georgia State University
Shi X, Ibrahim JG, Lieberman J, Styner M, Li Y, Zhu H (2011) Two-stage empirical likelihood for longitudinal neuroimaging data. Ann Appl Stat 5:1132–1159
Subramanian S (2007) Censored median regression and profile empirical likelihood. Stat Methodol 4(4):493–503
Sun J, Fang HB (2003) A nonparametric test for panel count data. Biometrika 90:199–208
Sun J, Kalbfleisch JD (1995) Estimation of the mean function of point processes based on panel count data. Stat Sin 5:279–289
Sun J, Tong X, He X (2007) Regression analysis of panel count data with dependent observation times. Biometrics 63:1053–1059
Sun J, Wei LJ (2000) Regression analysis of panel count data with covariate-dependent observation and censoring times. J R Stat Soc Ser B 62:293–302
Sun J, Zhao X (2013) Statistical analysis of panel count data. Springer, New York
Sun Y, Sundaram R, Zhao Y (2009) Empirical likelihood inference for the Cox model with time-dependent coefficients via local partial likelihood. Scand J Stat 36(3):444–462
Thall PF, Lachin JM (1988) Analysis of recurrent events: nonparametric methods for random-interval count data. J Am Stat Assoc 83:339–347
Therneau TM, Grambsch PM (1990) Martingale-based residuals for survival models. Biometrica 77(1):147–160
Wang S, Qian L, Carrol RJ (2010) Generalized empirical likelihood methods for analyzing longitudinal data. Biometrika 97(1):79–93
Wellner JA, Zhang Y (2000) Two estimators of the mean of a counting process with panel count. Ann Stat 28:779–814
Yang H, Zhao Y (2012) New empirical likelihood inference for linear transformation models. J Stat Plan Inference 142:1659–1668
Yu W, Sun Y, Zheng M (2011) Empirical likelihood method for linear transformation models. Ann Inst Stat Math 63:331–346
Yu W, Zhao Z, Zheng M (2012) Empirical likelihood methods based on influence functions. Stat Interface 5:355–366
Yu X, Zhao Y (2019) Jackknife empirical likelihood inference for the accelerated failure time model. TEST 28:269–288
Zhang H, Zhao H, Sun J, Wang D, Kim K (2013) Regression analysis of multivariate panel count data with an informative observation process. J Multivar Anal 119:71–80
Zhang Y (2002) A semiparametric pseudolikelihood estimation method for panel count data. Biometrika 89:39–48
Zhang Y (2006) Nonparametric \(k\)-sample tests with panel count data. Biometrika 93:777–790
Zhang Z, Zhao Y (2013) Empirical likelihood for linear transformation models with interval-censored failure time data. J Multivar Anal 116:398–409
Zhao H, Li Y, Sun J (2013) Analyzing panel count data with a dependent observation process and a terminal event. Can J Stat 41(1):174–191
Zhao X, Tong X, Sun J (2013) Robust estimation for panel count data with informative observation times. Comput Stat Data Anal 57:33–40
Zhao Y (2011) Empirical likelihood inference for the accelerated failure time model. Stat Probab Lett 81(5):603–610
Zhao Y, Jinnah A (2012) Inference for Cox’s regression models via adjusted empirical likelihood. Comput Stat 27:1–12
Zhao Y, Huang Y (2007) Test-based interval estimation under the accelerated failure time model. Commun Stat 36(3):593–605
Zhao Y, Yang S (2012) Empirical likelihood confidence intervals for regression parameters of the survival rate. J Nonparametr Stat 24(1):59–70
Acknowledgements
The authors would like to thank the Editor-in-Chief, Professor Carsten Jentsch, an associate editor, and two reviewers for the thoughtful comments and insightful suggestions, which significantly enhanced the quality and presentation of the manuscript. Dr. Faysal Satter acknowledges the financial support from the Georgia State University internal dissertation grant. Dr. Yichuan Zhao is very grateful to the support from NSF Grant (DMS-2317533) and the Simons Foundation Grant (grant number: 638679). Dr. Ni Li is also grateful to the grant support from Natural Science Foundation of Hainan Province (2019RC176).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix: Proofs of theorems
Appendix: Proofs of theorems
We assume the following regularity conditions throughout the paper, which are similar to Li et al. (2010).
-
(C.1)
The covariate vector X(t) and the weight function W(t) are bounded for all \(t \in [0,\tau ]\).
-
(C.2)
The function g is twice continuously differentiable.
-
(C.3)
The function W(t) converges uniformly to a deterministic function w(t) for all \( t \in [0, \tau ] \) a.s.
-
(C.4)
There exists a \( \tau >0 \) such that \( Pr(C_i \ge \tau ) > 0, i = 1, \cdots , n. \)
-
(C.5)
The matrix \(\Sigma \) is positive definite for all \(t \in [0,\tau ]\).
-
(C.6)
\( A_*=E\left[ \int _0^{\tau } \{ Z(t) - {\bar{z}}(t;\gamma _0)\}^{\otimes 2} \Delta _i(t) e^{\gamma _0' Z(t)} \lambda _0(t) dt \right] \) is positive definite, where E is the expectation.
Lemma A.1
Assume that the conditions (C.1)–(C.6) hold. If \( \beta _0= (\beta _{10}', \beta _{20}')' \) are the true values of the parameters, then
Proof
Note that
and
Since \( {\hat{\gamma }}\) is obtained by solving \( V(\gamma )=0 \) at \( \gamma ={\hat{\gamma }} \), we have (cf. pp. 148–149, Therneau and Grambsch (1990))
Some simple algebra leads to
As shown in the Appendix of Li et al. (2010) that \( n^{-1/2} U(\beta _{0}; {\hat{\gamma }}) \) converges to a zero mean Gaussian distribution with covariance matrix \( \Sigma \), we have
\(\square \)
Lemma A.2
Assume that the conditions (C.1)-(C.6) hold. If \( \beta _0= (\beta _{10}', \beta _{20}')' \) are the true values of the parameters, then
Proof
It can be shown that
where
Under conditions (C.1) and (C.3), \( || \epsilon _{i1}||=o_p(1), || \epsilon _{i2}||=o_p(1), || \epsilon _{i5}||=o_p(1)\) and \(|| \epsilon _{i6}||=o_p(1) \) hold. Note that (cf. pp. 1103–1104, Andersen and Gill (1982))
Then under conditions (C.1)-(C.3) and by the consistency of \({\hat{\mu }}_0(t) \) to \( {\mu }_0(t) \), we have
In the term \( E_1,\) for each i, \( w(t) X_i(t) g\{{\mu }_0(t)e^{\beta _0'X_i(t)}\} \Delta _i(t) e^{\gamma _0'Z_i(t)} \Big / s^{(0)}(t; \gamma _0) \) is predictable and finite, and \( M_i^* (t) \) is a martingale. Then
is a martingale integral and converges to zero in probability. Also, similar to p. 300 of Fleming and Harrington (1991), the Taylor series expansion of \({\hat{\Lambda }}_0(t;{\hat{\gamma }})-{\hat{\Lambda }}_0(t;\gamma _0) \) about \( \gamma =\gamma _0 \) in the term \( E_2\) can be written as
where \( \gamma ^* \) is on the line segment between \( {\hat{\gamma }} \) and \( \gamma _0 \) and H is defined as
Following Lin et al. (2000), it can be shown that \( H(t;\gamma _0) \) converges to the following function a.e.
uniformly in t and
Therefore, \({\hat{\Lambda }}_0(t;{\hat{\gamma }})-{\hat{\Lambda }}_0(t;\gamma _0) \) in the term \( E_2\) is tight and therefore equal to
Hence,
where \( \int _0^{\tau } w(t) X_i(t) g\{{\mu }_0(t) e^{\beta _0'X_i(t)}\} \Delta _i(t) e^{\gamma _0'Z_i(t)} [h'(t;\gamma _0) A_*^{-1} \sum _{i=1}^n \frac{1}{n} \{ Z_i(t)- {\bar{z}}(t;\gamma _0)\} d M_i^*(t)] \) is a martingale integral and converges to zero in probability. Thus, \( \epsilon _{i3} = o_p(1).\) Using similar arguments, it can be shown that \( \epsilon _{i4}=o_p(1) \) and \( \epsilon _{i7}=o_p(1).\) Therefore,
Let \( {\hat{Q}}_n = {n^{-1}} \sum _{i=1}^n U_{ni}(\beta _0;{\hat{\gamma }}) (U_{0i}(\beta _0;{\hat{\gamma }}))'\) and \( {Q_n} = {n^{-1}} \sum _{i=1}^n {U_{0i}}(\beta _{0};\gamma _0) {U_{0i}}(\beta _{0};\gamma _0))'\). For any \( c \in R^p\), the following decomposition holds:
Both \( I_1 \) and \( I_2 \) are \( o_p(1) \) by (A.1). As a result,
\(\square \)
Proof of Theorem 1
Following Owen (1988, 1990), we can prove the theorem similarly. From Lemma A.1, we have \( \underset{n \rightarrow \infty }{ \text{ lim }} {\hat{Q}}_n = \Sigma .\) As \( U_i(\beta _{0}; \gamma _0) \) are i.i.d. r.v. Then combining Owen (2001) and (A.1), we have
Combining Lemma A.1, Eq. (8) and Taylor expansion of \(l(\beta _{0})\), one can prove Theorem 1. The details are omitted here. \(\square \)
Proof of Theorem 2
The proof is similar to arguments from Qin and Lawless (1994), Yu et al. (2011), Yang and Zhao (2012), Zhang and Zhao (2013), and Yu and Zhao (2019). By Lemma A.1, we complete the proof of Theorem 2. The details are omitted here. \(\square \)
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Satter, F., Zhao, Y. & Li, N. Empirical likelihood inference for the panel count data with informative observation process. Stat Papers (2023). https://doi.org/10.1007/s00362-023-01506-0
Published:
DOI: https://doi.org/10.1007/s00362-023-01506-0