Sum of coefficient constraint and ARMA

[fboehlandt]

Hi everyone,
how can i impose the sum of coefficient constraint = 1 for a univariate ARMA model? The following yields 'AR not defined':
y = c(1) + (c(2) / (c(2) + c(3))) * ar(1) + (c(3) / (c(2) + c(3))) * ma(1)
Any help is greatly appreciated

[EViews Gareth]

ARMA terms are entered in an equation by expression with the following syntax:

Code: Select all

y=c(1)*x + c(2)*z + [ar(1)=c(3)]

[fboehlandt]

Hello Gareth,
thanks for the reply. I suppose I can impose restrictions on the coefficients for an AR[] process using the lagged series. However, I am still struggling imposing the constraints for moving-average process since the error series is not observable. The following yields a syntax error:

y = c(1) + [ar(1)=c(2)/(c(2) + c(3)), ma(1)=c(3)/(c(2) + c(3))]

I am trying to describe an ARMA(p,q) process where the coefficient estimates c(2) + c(3) + c(4) +... = 1