摘要 |
<p><P>PROBLEM TO BE SOLVED: To provide an accurate parameter estimation method by introducing a new technique to a bootstrap method not always excellent in convergence conventionally, and by reducing a calculation time in an ARMA model to which a control input is not applied. <P>SOLUTION: When a time series comprising N pieces of data is identified by a (p, q) succeeding ARMA model by use of a computer equipped with a storage part and a calculation part, äx<SB>n</SB>}; a sample time series (n=1, 2, .., N..) of a steady ergodic normal process x(t), and äe<SB>n</SB>}; the normalization white noise of an average value 0 and dispersionσ<SB>e</SB><SP>2</SP>. Regarding an autoregressive (AR) part, a parameter is estimated by use of a tertiary cumulant. <P>COPYRIGHT: (C)2006,JPO&NCIPI</p> |