Abstract
In this study, a pseudo-maximization by parts method is introduced by developing the maximization by parts algorithm for the parameter estimation of pseudo-copula regression models. Sub- and main score equations are obtained from the pairwise log-likelihood function and solved by the proposed iterative algorithm. The pseudo-maximization by parts algorithm is an iterative algorithm to avoid having to calculate the second-order derivative of the full log-likelihood function as maximization by parts algorithm. Instead of the Gaussian copula function in maximization by parts algorithm, the pseudo-Gaussian copula function is included in the new algorithm. The mean square errors of the estimators found by the maximization by parts algorithm and the pseudo-maximization by parts algorithm are compared using real Turkish comprehensive insurance data taken from the Turkish Insurance Information and Monitoring Center for the year 2017, and it is notable that the proposed algorithm provided better results in terms of having lower errors.
Similar content being viewed by others
References
Brigo D, Pallavicini A, Torresetti R (2010) Credit models and the crisis: a journey into CDOs, copulas, correlations and dynamic models. Wiley, New York
Czado C, Kastenmeier R, Brechmann EC, Min A (2012) A mixed copula model for insurance claims and claim sizes. Scand Actuar J 4:278–305
Erdemir ÖG (2020) Modified pseudo-copula regression model. Ph.d. thesis, Hacettepe University
Erdemir ÖK, Sucu M (2022) A modified pseudo-copula regression model for risk groups with various dependency levels. J Stat Comput Simul 92(5):1092–1112
Fang Y, Madsen L (2013) Modified Gaussian pseudo-copula: application in insurance and finance. Insur Math Econ 53(1):292–301
Fermanian JD (2005) Goodness-of-fit tests for copulas. J Multivar Anal 95(1):119–152
Fermanian JD, Wegkamp M (2012) Time-dependent copulas. J Multivar Anal 110:19–29
Genest C, Rémillard B, Beaudoin D (2009) Goodness-of-fit tests for copulas: a review and a power study. Insur Math Econ 44(2):199–213
Nelsen RB (2007) An introduction to copulas. Springer, Berlin
Patton AJ (2006) Estimation of multivariate models for time series of possibly different lengths. J Appl Econ 21(2):147–173
R Core Team (2020) R: a language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. http://www.R-project.org/
Sklar M (1959) Fonctions de répartition à n dimensions et leurs marges. Publications De L’institut De Statistique De L’université De Paris 8:229–231
Song PXK (2007) Correlated data analysis: modeling, analytics, and applications. Springer, Berlin
Song PXK, Fan Y, Kalbfleisch JD (2005) Maximization by parts in likelihood inference. J Am Stat Assoc 100(472):1145–1158
Song PXK, Li M, Yuan Y (2009) Joint regression analysis of correlated data using Gaussian copulas. Biometrics 65(1):60–68
White H (1996) Estimation, inference and specification analysis (No. s). Cambridge University Press, Cambridge
Weiß G (2011) Copula parameter estimation by maximum-likelihood and minimum-distance estimators: a simulation study. Comput Stat 26(1):31–54
Zhang J, Gao K, Li Y, Zhang Q (2022) Maximum likelihood estimation methods for copula models. Comput Econ 60(1):99–124
Funding
The author declares that no funds, grants, or other support were received during the preparation of this manuscript.
Author information
Authors and Affiliations
Corresponding author
Ethics declarations
Conflict of interest
The author has no relevant financial or non-financial interests to disclose. Author ÖK. Erdemir declares that she has no conflict of interest.
Ethical approval
This article does not contain any studies with human participants or animals performed by the author.
Availability of data and material
The data were provided by Turkish Insurance Information and Monitoring Center with restricted use. I am not allowed to share with third parties.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix
Appendix
Let \(\frac{\partial {D}_{{\rho }_{12}^{*}}\left({F}_{{X}_{1}}\left({x}_{i1}\right),{F}_{{X}_{2}}\left({x}_{i2}\right)\right)}{\partial{\varvec{\beta}}}\) be multiplied by \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}\) and \(\frac{\partial {D}_{{\rho }_{12}^{*}}\left({F}_{{X}_{1}}\left({x}_{i1}\right),{F}_{{X}_{2}}\left({x}_{i2}-1\right)\right)}{\partial{\varvec{\beta}}}\) be multiplied by \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}-1\right)}{\partial {F}_{{X}_{2}}\left({x}_{i2}-1\right)}\).
Partial integrals \(\frac{\partial {D}_{{\rho }_{12}^{*}}\left({F}_{{X}_{1}}\left({x}_{i1}\right),{F}_{{X}_{2}}\left({x}_{i2}\right)\right)}{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}\), \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}{\partial{\varvec{\beta}}}\), \(\frac{\partial {D}_{{\rho }_{12}^{*}}\left({F}_{{X}_{1}}\left({x}_{i1}\right),{F}_{{X}_{2}}\left({x}_{i2}-1\right)\right)}{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}\) and \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}-1\right)}{\partial{\varvec{\beta}}}\) are expressed one by one below.
Let \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}{\partial{\varvec{\beta}}}\) be multiplied by \(\frac{\partial {\lambda }_{i}}{\partial {\lambda }_{i}}\). For a Poisson GLM \({\lambda }_{i}={e}^{{{\varvec{z}}}_{{\varvec{i}}2}^{\boldsymbol{^{\prime}}}{\varvec{\beta}}}\), \(\frac{\partial {\lambda }_{i}}{\partial{\varvec{\beta}}}={{\varvec{z}}}_{{\varvec{i}}2}^{\boldsymbol{^{\prime}}}{e}^{{{\varvec{z}}}_{{\varvec{i}}2}^{\boldsymbol{^{\prime}}}{\varvec{\beta}}}={{\varvec{z}}}_{{\varvec{i}}2}^{\boldsymbol{^{\prime}}}{\lambda }_{i}\) is written.
\(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}{\partial {\lambda }_{i}}\) is expressed as \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}{\partial {\lambda }_{i}}=\frac{\partial }{\partial {\lambda }_{i}}\sum_{k=0}^{{x}_{i2}}\frac{1}{k!}{{\lambda }_{i}}^{k}{e}^{-{\lambda }_{i}}=\sum_{k=0}^{{x}_{i2}}\frac{\partial }{\partial {\lambda }_{i}}\left(\frac{1}{k!}{{\lambda }_{i}}^{k}{e}^{-{\lambda }_{i}}\right)\).
Since \(0\le k\le {x}_{i2}\), when deriving at the point k = 0, the expression \(\sum_{k=0}^{{x}_{i2}}\frac{1}{(k-1)!}\dots .\) is encountered. The range is separated as \(k=0\) and \(1\le k\le ,\) since the factorial operation is undefined for negative numbers.
\(\frac{\partial {D}_{{\rho }_{12}^{*}}\left({F}_{{X}_{1}}\left({x}_{i1}\right),{F}_{{X}_{2}}\left({x}_{i2}-1\right)\right)}{\partial {F}_{{X}_{2}}\left({x}_{i2}\right)}\) and \(\frac{\partial {F}_{{X}_{2}}\left({x}_{i2}-1\right)}{\partial{\varvec{\beta}}}\) are the discrete forms of Eqs. (1) and (2), respectively.
\(\frac{{\partial F_{{X_{2} }} \left( {x_{i2} - 1} \right)}}{{\partial {\varvec{\beta}}}} = \frac{{\partial F_{{X_{2} }} \left( {x_{i2} - 1} \right)}}{{\partial \lambda_{i} }}\frac{{\partial \lambda_{i} }}{{\partial {\varvec{\beta}}}} = \frac{{\partial F_{{X_{2} }} \left( {x_{i2} - 1} \right)}}{{\partial \lambda_{i} }}{\varvec{z}}_{{{\varvec{i}}2}}^{\user2{^{\prime}}} \lambda_{i}\)
All components for the solution of the second main score equation are obtained by Eqs. (1)–(4), and \(\frac{\partial {l}_\mathrm{dep}\left({{\varvec{\theta}}}_{1},{{\varvec{\theta}}}_{2}\right)}{\partial{\varvec{\beta}}}\) is written in Eq. (5) with the definition of \({d}_{{\rho }_{12}^{*}}\left(.,.\right)\).
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
Erdemir, Ö.K. An adjusted iterative algorithmic approach for maximum likelihood estimation of the pseudo-copula regression: P-MBP. Soft Comput 27, 11227–11241 (2023). https://doi.org/10.1007/s00500-023-08722-8
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00500-023-08722-8