Higher-order lag specification in sspace
Posted: Fri Mar 18, 2022 4:12 am
Dear all,
I am having a hard time writing modular code to allow lagged variables in a state-space model in EViews.
Suppose that there are many signals and two common factors with an AR(2) structure:
Then, I define the state variables an an AR(2) process:
This syntax comes directly from the EViews manual (the ARMAX(2,3) example). Is there a better way to do it? I am currently doing model selection, and having to define extra state variables, adding or removing lines, instead of just using higher-order lags in the state equation is quite cumbersome...
Maybe there are plans for the next release of EViews to allow syntax like
this?
Yours sincerely,
Andreï V. Kostyrka
I am having a hard time writing modular code to allow lagged variables in a state-space model in EViews.
Suppose that there are many signals and two common factors with an AR(2) structure:
Code: Select all
sspace emp
emp.append @signal growthrate01 = c(101)*SV1 + c(201)*SV2 + U01
...
emp.append @signal growthrate08 = c(108)*SV1 + c(208)*SV2 + U08
Then, I define the state variables an an AR(2) process:
Code: Select all
emp.append @state SV1 = C(11) * SV1(-1) + C(12) * SV1_1(-1) + ESV1
emp.append @state SV1_1 = SV1(-1)
emp.append @state SV2 = C(21) * SV2(-1) + C(22) * SV2_1(-1) + ESV2
emp.append @state SV2_1 = SV2(-1)
This syntax comes directly from the EViews manual (the ARMAX(2,3) example). Is there a better way to do it? I am currently doing model selection, and having to define extra state variables, adding or removing lines, instead of just using higher-order lags in the state equation is quite cumbersome...
Maybe there are plans for the next release of EViews to allow syntax like
Code: Select all
emp.append @state SV1 = C(11) * SV1(-1) + C(12) * SV1(-2) + ESV1
emp.append @state SV2 = C(21) * SV2(-1) + C(22) * SV2(-2) + ESV2
this?
Yours sincerely,
Andreï V. Kostyrka