ESTIMATION OF IMPULSE RESPONSE FUNCTION BY ITERATIVELY APPLIED MULTIPLE REGRESSION

In this paper the author uses iterative multiple regression and backward elimination process to determine the impulse response function coefficients of a given pair of input/output process. The computer-based solution is done with the help of a Pascal program, which organises the selection of the input variables with increasing time lag for the iterative solutions. The truncation point is determined by using the error square contributions of the individual input variables. The method proves stable in both numerical and statistical sense. There was no instability observed up to 80 input variables. KEY WORDS: Iterative multiple regression, impulse response function, error square contribution, time series analysis, transfer function model Global Jnl of Mathematical Sciences Vol. 3(1) 2004: 47-5