A user-knowledge-based variable selection method for limited information maximum likelihood using principal components. (English) Zbl 0875.62252
Summary: As analysis or prediction is made in detail, a simultaneous equation system is enlarged. Since a researcher usually has no perfect knowledge and/or precise data about the system, he estimates each equation of the system to some extent by trial and error. It becomes time-consuming, laborious and costly to estimate the whole system. If the number of predetermined variables exceeds the sample size or strong multicollinearity occurs among excluded predetermined variables, the system cannot be estimated with the limited information maximum likelihood (LIML). We modify LIML by using some effectively-influential principal components of excluded predetermined variables and propose a practical variable selection method to solve the j-th best subset problem in one computer-run by user’s professional knowledge regarding the research in question as well as statistical and data-analytic criteria. The proposed method is embodied in the Researcher System OEPP.
MSC:
| 62H25 | Factor analysis and principal components; correspondence analysis |
Keywords:
Limited information maximum likelihood; Principal component analysis; The j-th best subset problem; User-knowledge-based variable selection method; Scientific variable classification; Econometric variable classification; Researcher System OEPPReferences:
| [1] | Akaike, H.: Information theory and an extension of the maximum likelihood principle. Proc. of 2nd inter. Symp. inform. Theor., 178-196 (1973) · Zbl 0283.62006 |
| [2] | Almon, S.: The distributed lag between capital approximations and expenditures. Econometrica 30, 178-196 (1965) |
| [3] | Anderson, T. W.; Rubin, H.: Estimation of the parameters of a single equation in a complete system of stochastic equations. Ann. math. Statist. 20, 46-63 (1949) · Zbl 0033.08002 |
| [4] | Anderson, T. W.; Rubin, H.: The asymptotic properties of estimators of the parameters of a single equation in a complete system of stochastic equations. Ann. math. Statist. 21, 570-581 (1950) · Zbl 0039.36002 |
| [5] | Basmann, R. L.: On finite sample distributions of generalized classical linear identifiability test. J. amer. Statist. assoc. 55, No. 292, 650-659 (1960) · Zbl 0096.35401 |
| [6] | Durbin, J.: Testing for serial correlation in systems of simultaneous regression equations. Biometrika 44, No. 3, 370-377 (1975) · Zbl 0081.36301 |
| [7] | Durbin, J.; Watson, G. S.: Testing for serial correlation in least squares regression, I. Biometrika 37, 409-428 (1950) · Zbl 0039.35803 |
| [8] | Durbin, J.; Watson, G. S.: Testing for serial correlation in least squares regression, II. Biometrika 38, 159-178 (1951) · Zbl 0042.38201 |
| [9] | Hood, W. C.; Koopmans, T. C.: Studies in econometric methods. (1953) · Zbl 0053.27905 |
| [10] | Liu, T. C.; Breen, W. J.: The covariance matrix of the limited information estimator and the identification test. Econometrica 37, No. 2, 222-227 (1969) |
| [11] | Onishi, H.: A variable selection procedure for econometric models. Comput. statist. & data anal. 1, No. 2, 85-95 (1983) · Zbl 0555.62094 |
| [12] | Onishi, H.: Researcher system OEPP for econometrics: I and II. (1991) |
| [13] | Shiller, R. J.: A distributed lag estimator derived from smoothness priors. Econometrica 41, No. 4, 775-788 (1973) · Zbl 0305.62070 |
This reference list is based on information provided by the publisher or from digital mathematics libraries. Its items are heuristically matched to zbMATH identifiers and may contain data conversion errors. In some cases that data have been complemented/enhanced by data from zbMATH Open. This attempts to reflect the references listed in the original paper as accurately as possible without claiming completeness or a perfect matching.
