Estimation of factor models by realization-based and approximation methods
Wolfgang Scherrer, Christiaan Heij · 1998
In this paper we discuss two methods for the estimation of linear dynamic factor models. The first method is behavioural in nature and consists of the least squares approximation of the observed data by means of a linear system. The second method is based on the statistical concept of principal components and uses subspace ideas from approximate realization theory. The two methods are compared by means of simulated data. Support by the Austrian "Fonds zur Forderung der wissenschaftlichen Forschung" Project P-11213-MAT is gratefully acknowledged. 1 1 Introduction The essential ingredients of system identification consist of data, models, and methods to identify models from data. In this paper we restrict the attention to time series data and linear system models and we focus on identification methods within this setting. These methods can be classified in terms of the following specifications of the identification problem : a. symmetric treatment of variables or not ? b. equations...