![]() | Only 14 pages are availabe for public view |
Abstract It is well known that time series inalysis u5ually becomes much more complicated whi a it L account of superimposed observational ea suppocd the superimposed error as disc (1944) .ome times it is impossibl to obs which is described by a linear siationar. take into account the error of tite obser be for example the results of th instn [n this work we shall cone der the proce&..es xc..t) which are, the ac oregr& movin average MA(q) process and mixed t moving average ARMA. (p,q) proce& .. In e suppose it is Gaussian process wLth mesi examine the likelihood estimatic of th’. these kinds of processes with er’or, fo conditions (Eee L1) the likeli ±ood eqw.tion has a solution which converges in proLabilit to the true value of the parameter. It wil] be assumed that the error is a sequence of Gaussian random variables with mean zer and variance We assume Lnat the two processes X(t) , t) are independent. necessary to or. The problem ssed by Kandell ye the process nodel. V’e must tions which may nts. he originaly ‘ze process ARC;) oregressive y case we shall crc. We shall .arameters of rider certain |