Page 1 of 1

Forecasting Univariate Time Series

Posted: Fri Apr 15, 2011 4:52 pm
by john_stanley
Hi,

I need some clarifications on making forecast on HPI. I'm using Eviews 6.

My data is HPI. I've transformed it to log to LHPI using and it is stationary at 1st difference.

The correlogram of D(LHPI) shows that there is siginificant lags at lag 1,2 and 7 of PACF and lag 1 and 7 of ACF

then I regress the data ; LS D(LHPPI) AR(1) AR(2) AR(7) MA(1) MA(7) C , and the result shows that p-value at AR(2) is >5%, others were OK.

can I just regress with LS D(LHPPI) AR(1) MA(7) C , as all of the p-value were all 0 ? I need some clarifications on the problem.

can I just picking any of the lags significant from the correlogram for LS regression such as AR(1) and MA(7) only?

hope you may help me. thanks in advance.

Re: Forecasting Univariate Time Series

Posted: Tue Apr 19, 2011 10:31 pm
by john_stanley
still no answer?