ESTIMASI PARAMETER MODEL TIME SERIES REGRESSION DENGAN NOISE ARMA(p) DAN ARCH(r)-MEAN MENGGUNAKAN METODE MAXIMUM LIKELIHOOD
Abstract: In regression analysis
to obtain a good model and valid it must be satisfied that the assumptions residuals
is Normal distributed with mean zero and constant variance (homoscedastic), and
residual nonautocorrelation. In addition, the predictor variables are
independent with other predictor variables (nonmulticollinearity). However, in
reality the financial data of these assumptions are rarely met, so it is
necessary to residual remodeling. This modeling is called the Time Series
Regression model ARCH-M models
(Autoregressive Conditional
Heteroscedastic in Mean) is one of the time series models which
have a non-constant variance error. ARCH-M model is obtained
by extending the basic framework
of ARCH models which calculating mean series dependent mean variance error. So
in this model, variant residual inserted into the equations mean.
This final project had purpose to establish and estimate the model of
Time Series Regression withAR(p) noise
and ARCH (r)-Mean using maximum likelihood method. Then applied to the
data volume nd price of EURO exchange rate against the USD 30-minute time frame
in the type of ECN (Electronic CummunicationNetwork) at Alpari-US companies on
18 April 2012 until 20 April 2012.
Based on the application of the data using Eviews software obtained that
best models are the model of Time Series Regression with noise AR(2) and ARCH
(1)-Mean so the model is𝑦𝑡 =161770.5𝑥𝑡1+615494.8𝑥𝑡2−921821.1𝑥𝑡3+144593.5𝑥𝑡4+0.383620𝜀𝑡−1+0.270385𝜀𝑡−2+1.412238�ℎ𝑡+𝑣𝑡ℎ𝑡
=91765.19+0.414174𝑣𝑡−1 With AIC and SBC values
respectively are 14.76345 and 14.95620.
Penulis: Athoillah, Sediono,
Suliyanto
Kode Jurnal: jpmatematikadd130071
