Page 1 of 1
%mvgavge
Posted: Tue Feb 01, 2011 4:54 pm
by ivory4
It is explained a little bit in VAdd and Appendix B.
Also used in extracting
Code: Select all
boxjenk(....define=xx) xxx
com mapoly = %eqnlagpoly(xx, %mvgavge)
What is extracted using %mvgavge?
I use SET a / =%mvgavge. It shows nothing.
If I need to calculate (1+%beta(3))/(1-%beta(2)) after fitting ARIMA(1,1,1), is it correct
Code: Select all
com mapoly = %eqnlagpoly(xx, %mvgavge)
com arpoly = %eqnlagpoly(xx, y)
com com psi1=%polyvalue(mapoly,1)/%polyvalue(arpoly,1)
But it seems not working
Re: %mvgavge
Posted: Tue Feb 01, 2011 7:22 pm
by TomDoan
ivory4 wrote:It is explained a little bit in VAdd and Appendix B.
Also used in extracting
Code: Select all
boxjenk(....define=xx) xxx
com mapoly = %eqnlagpoly(xx, %mvgavge)
What is extracted using %mvgavge?
I use SET a / =%mvgavge. It shows nothing.
It's not a series itself, but a special descriptor for the residuals in certain types of models.
ivory4 wrote:
If I need to calculate (1+%beta(3))/(1-%beta(2)) after fitting ARIMA(1,1,1), is it correct
Code: Select all
com mapoly = %eqnlagpoly(xx, %mvgavge)
com arpoly = %eqnlagpoly(xx, y)
com com psi1=%polyvalue(mapoly,1)/%polyvalue(arpoly,1)
But it seems not working
That should be correct:
Input:
Code: Select all
all 100
seed 13430
set y = %ran(1.0)
boxjenk(ar=1,ma=1,constant,define=bbeq) y
compute mapoly=%eqnlagpoly(bbeq,%mvgavge)
compute arpoly=%eqnlagpoly(bbeq,y)
compute psi1=%polyvalue(mapoly,1)/%polyvalue(arpoly,1)
disp psi1
disp 2.117747/1.749825165
Output:
1. CONSTANT 0.006577036 0.088758699 0.07410 0.94108486
2. AR{1} -0.749825165 0.047736721 -15.70751 0.00000000
3. MA{1} 1.117747488 0.043617580 25.62608 0.00000000
1.21026
1.21026
Re: %mvgavge
Posted: Tue Feb 01, 2011 11:18 pm
by ivory4
I am not sure what is wrong in this case
Code: Select all
calendar(q) 1979:1
open data japan79rgdp.xls
data(format=xls,org=columns) / rgdp
set rgdp / = 100.0*log(rgdp)
boxjenk(diffs=1,ar=1,ma=1,maxl,$
constant,define=bbeq) rgdp
compute mapoly=%eqnlagpoly(bbeq,%mvgavge)
compute arpoly=%eqnlagpoly(bbeq,rgdp)
display %polyvalue(mapoly,1)/%polyvalue(arpoly,1)
display (1+%beta(2))/(1-%beta(3))
Results:
Variable Coeff Std Error T-Stat Signif
********************************************************************************
1. CONSTANT 0.764442812 0.267839078 2.85411 0.00586058
2. AR{1} 0.918498499 0.154688083 5.93775 0.00000014
3. MA{1} -0.773436855 0.223787014 -3.45613 0.00099491
NA
1.08180
Re: %mvgavge
Posted: Wed Feb 02, 2011 7:03 am
by TomDoan
You're estimating the equation with the DIFFS option. The AR polynomial will include the 1-L operator, so your denominator will be zero. You'll have to estimate the equation on the differenced data, rather than the non-stationary data.