Computing all roots of the likelihood equations of seemingly unrelated regressions

Seemingly unrelated regressions are statistical regression models based on the Gaussian distribution. They are popular in econometrics but also arise in graphical modeling of multivariate dependencies. In maximum likelihood estimation, the parameters of the model are estimated by maximizing the likelihood function, which maps the parameters to the likelihood of observing the given data. By transforming this optimization problem into a polynomial optimization problem, it was recently shown that the likelihood function of a simple bivariate seemingly unrelated regressions model may have several stationary points. Thus local maxima may complicate maximum likelihood estimation. In this paper, we study several more complicated seemingly unrelated regression models, and show how all stationary points of the likelihood function can be computed using algebraic geometry.

[1]  L. Telser,et al.  Iterative Estimation of a Set of Linear Regression Equations , 1964 .

[2]  A. Goldberger.  A course in econometrics , 1991 .

[3]  David A. Cox,et al.  Ideals, Varieties, and Algorithms , 1997 .

[4]  D. Madigan,et al.  Alternative Markov Properties for Chain Graphs , 2001 .

[5]  T. Richardson,et al.  Multimodality of the likelihood in the bivariate seemingly unrelated regressions model , 2004 .

[6]  W. Ames.  Algorithms (2nd edition) , 1990 .

[7]  M. Drton,et al.  Conditional independence models for seemingly unrelated regressions with incomplete data , 2006 .

[8]  J. Rochon.  Accounting for Covariates Observed Post Randomization for Discrete and Continuous Repeated Measures Data , 1996 .

[9]  Pablo A. Parrilo,et al.  Minimizing Polynomial Functions , 2001, Algorithmic and Quantitative Aspects of Real Algebraic Geometry in Mathematics and Computer Science.

[10]  J Rochon,et al.  Analyzing bivariate repeated measures for discrete and continuous outcome variables. , 1996, Biometrics.

[11]  S. Basu,et al.  Algorithmic and Quantitative Real Algebraic Geometry , 2003 .

[12]  Jan Kmenta,et al.  A General Procedure for Obtaining Maximum Likelihood Estimates in Generalized Regression Models , 1974 .

[13]  B. Sturmfels.  SOLVING SYSTEMS OF POLYNOMIAL EQUATIONS , 2002 .

[14]  P. Spirtes,et al.  Ancestral graph Markov models , 2002 .

[15]  M. Perlman,et al.  Normal linear models with lattice conditional independence restrictions , 1994 .

[16]  W. N. Venables,et al.  An extension of the growth curve model , 1988 .

[17]  David A. Cox,et al.  Using Algebraic Geometry , 1998 .

[18]  A. Zellner.  An Efficient Method of Estimating Seemingly Unrelated Regressions and Tests for Aggregation Bias , 1962 .