Abstract
A robust Bayesian model for seemingly unrelated regression is proposed. By using heavy-tailed distributions for the likelihood, robustness in the response variable is attained. In addition, this robust procedure is combined with a diagnostic approach to identify observations that are far from the bulk of the data in the multivariate space spanned by all variables. The most distant observations are downweighted to reduce the effect of leverage points. The resulting robust Bayesian model can be interpreted as a heteroscedastic seemingly unrelated regression model. Robust Bayesian estimates are obtained by a Markov Chain Monte Carlo approach. Complications by using a heavy-tailed error distribution are resolved efficiently by representing these distributions as a scale mixture of normal distributions. Monte Carlo simulation experiments confirm that the proposed model outperforms its traditional Bayesian counterpart when the data are contaminated in the response and/or the input variables. The method is demonstrated on a real dataset.
Similar content being viewed by others
References
Agostinelli C, Greco L (2013) A weighted strategy to handle likelihood uncertainty in Bayesian inference. Comput Stat 28:319–339
Ando T (2011) Bayesian variable selection for the seemingly unrelated regression models with a large number of predictors. J Jpn Stat Soc 41:187–203
Ando T, Zellner A (2010) Hierarchical Bayesian analysis of the seemingly unrelated regression and simultaneous equations models using a combination of direct Monte Carlo and importance sampling techniques. Bayesian Anal 5:65–95
Andrade JAA, O’Hagan A (2006) Bayesian robustness modeling using regularly varying distributions. Bayesian Anal 1:169–188
Arslan O (2010) An alternative multivariate skew Laplace distribution: properties and estimation. Stat Pap 51:865–887
Bayarri M, Morales J (2003) Bayesian measures of surprise for outlier detection. J Stat Plan Inference 111:3–22
Benoit DF, Van Aelst S, Van den Poel D (2015) Outlier-robust Bayesian multinomial choice modeling. J Appl Econ 31:1445–1466
Berger J (1994) An overview of robust Bayesian analysis (with discussion). Test 3:5–124
Billio M, Casarin R, Rossini L (2016) Bayesian nonparametric sparse seemingly unrelated regression model (SUR). Working Papers 2016:20, Department of Economics, University of Venice “Ca’ Foscari”
Bilodeau M, Duchesne P (2000) Robust estimation of the SUR model. Can J Stat 28:277–288
Chib S, Greenberg E (1995) Hierarchical analysis of sur models with extensions to correlated serial errors and time-varying parameter models. J Econ 68:339–360
Choi HM, Hobert JP (2013) Analysis of MCMC algorithms for Bayesian linear regression with Laplace errors. J Multivar Anal 117:32–40
Farcomeni A, Greco L (2015) Robust methods for data reduction. Chapman & Hall-CRC, Boca Raton
García-Escudero L, Gordaliza A, Matrán C, Mayo-Iscar A (2011) Exploring the number of groups in robust model-based clustering. Stat Comput 21:585–599
Gelman A, Carlin JB, Stern H, Dunson D, Vehtari A, Rubin D (2015) Bayesian data analysis, 3rd edn. Chapman & Hall-CRC, Boca Raton
Geweke J (1993) Bayesian treatment of the independent student-t linear model. Appl Econ 8:19–40
Greco L, Racugno W, Ventura L (2008) Robust likelihood functions in Bayesian inference. J Stat Plan Inference 138:1258–1270
Grunfeld Y (1958) The Determinants of Corporate Investment. PhD thesis, Department of Economics, University of Chicago
Hubert M, Rousseeuw PJ, Van Aelst S (2008) High-breakdown robust multivariate methods. Stat Sci 23:92–119
Koenker R, Portnoy S (1990) M-estimation of multivariate regressions. J Am Stat Assoc 85:1060–1068
Kotz S, Kozubowski T, Podgorski K (2001) The Laplace distribution and generalizations: a revisit with applications to communications, economics, engineering, and finance. Springer, New York
Lavine M (1991) Sensitivity in Bayesian statistics: the prior and the likelihood. J Am Stat Assoc 86:396–399
Lavine M (1994) An approach to evaluating sensitivity in Bayesian regression analyses. J Stat Plan Inference 40:233–244
Maronna RA, Martin DR, Yohai VJ (2006) Robust statistics: theory and methods. Wiley, New York
Ng VM (2002) Robust Bayesian inference for seemingly unrelated regressions with elliptical errors. J Multivar Anal 83:409–414
Peña D, Zamar R, Yan G (2009) Bayesian likelihood robustness in linear models. J Stat Plan Inference 139:2196–2207
Peremans K, Van Aelst S (2018) Robust inference for seemingly unrelated regression models. J Multivar Anal 167:212–224
Sivaganesan S (1993) Robust Bayesian diagnostics. J Stat Plan Inference 35:171–188
Train K (2003) Discrete choice methods with simulation. Cambridge University Press, Cambridge
Verzilli CJ, Stallard N, Whittaker JC (2005) Bayesian modelling of multivariate quantitative traits using seemingly unrelated regressions. Genet Epidemiol 28:313–325
Watson J, Holmes C (2016) Approximate models and robust decisions. Stat Sci 31:465–489
Zellner A (1962) An efficient method of estimating seemingly unrelated regressions and tests for aggregation bias. J Am Stat Assoc 57:348–368
Zellner A (1996) An introduction to Bayesian inference in econometrics. Wiley, New York
Zellner A, Ando T (2010a) Bayesian and non-Bayesian analysis of the seemingly unrelated regression model with student-t errors, and its application for forecasting. Int J Forecast 26:413–434. Special Issue: Bayesian Forecasting in Economics
Zellner A, Ando T (2010b) A direct Monte Carlo approach for Bayesian analysis of the seemingly unrelated regression model. J Econ 159:33–45
Author information
Authors and Affiliations
Corresponding author
Additional information
This work was carried out using the Stevin Supercomputer Infrastructure at Ghent University, funded by Ghent University, the Hercules Foundation and the Flemish Government. The research of Stefan Van Aelst was supported by Project C16/15/068 of Internal Funds KU Leuven, CRoNoS COST Action IC1408, and IAP research network Grant Nr. P6/03 of the Belgian government (Belgian Science Policy)
Appendix
Appendix
Theorem 1
We follow the lines of the proof of Theorem 2 in Peña et al. (2009). This means that we have to show that the weighted observations \(z_i^\star \) are bounded which implies that the Kullback–Leibler divergence is bounded as well. Let Z denote the original dataset. After rearrangement of the observations, Z can be split into \(Z=\left( {\begin{matrix} Z_1 \\ Z_2 \end{matrix}}\right) \). \(Z_1\) contains the \(\lceil (1-\alpha ) n \rceil \) observations that remain uncontaminated in \(Z_\alpha \), while \(Z_2\) contains the remaining \([\alpha n]\) observations that are replaced by outliers in \(Z_\alpha \). The contaminated data \(Z_\alpha \) can similarly be split into \(Z_\alpha =\left( {\begin{matrix} Z_1 \\ Z_{2,\alpha } \end{matrix}}\right) \). Moreover, denote \(Z_\alpha ^\star = W\,Z_\alpha \) with \(W=\text {diag}(\omega _{1},\ldots ,\omega _{N})\) and \(\omega _{i}=\omega (d_i)\) with \(d_i=d_i(Z_\alpha )\) given by (7).
We first show that the constant a which is the \((1-\alpha )\) quantile of the distances \(d_{i}\) is bounded. Since \(\alpha <1/2\) and the coordinate-wise median T and the scatter matrix C both have breakdown point 1/2 under the conditions of the theorem, there exist bounds \(B_T(Z_1)\) and \(B_l(Z_1)\) such that \(\Vert T(Z_\alpha )\Vert \le B_T(Z_1)\Vert <\infty \) and \( 0<B_l(Z_1) \le \lambda _{\min }(C(Z_\alpha ))\le \lambda _{\max }(C(Z_\alpha )) \le B_u(Z_1) < \infty \) where \(\lambda _{\min }(C(Z_\alpha ))\) and \(\lambda _{\max }(C(Z_\alpha ))\) are the smallest/largest eigenvalue of \(C(Z_\alpha )\). For every \(z_{\alpha ,i}\) in \(Z_\alpha \) it then follows that
For observations \(z_{\alpha ,i}\) in \(Z_1\), we have that \(z_{\alpha ,i}=z_i\) and the corresponding distances \(d_i\) have a finite lower and upper bound that only depends on \(Z_1\). This implies immediately that the \((1-\alpha )\) quantile of the distances \(d_{i}\) in \(Z_\alpha \) is bounded (and positive) whatever the outliers are.
Now, consider \(z_i^\star = \omega _{i}z_{\alpha ,i}\). If \(d_i > a \), then we have that
which is bounded. If \(d_i \le a \), then \(\omega _{i}=1\) and thus \(d_i^\star = d_i \le a\). Hence, \(d_i^\star \) has a finite upper bound for all observations in \(Z_\alpha \) and thus also \(\Vert z_i^\star \Vert \) has a finite upper bound for the transformed observations.
Rights and permissions
About this article
Cite this article
Mbah, C., Peremans, K., Van Aelst, S. et al. Robust Bayesian seemingly unrelated regression model. Comput Stat 34, 1135–1157 (2019). https://doi.org/10.1007/s00180-018-0854-3
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00180-018-0854-3