ARMA exact ML

[hxh]

Hi all,

I have a question how can I estimate an ARMA(2,1) model in Eviews ?

Further I have calculated the asymptotic distribution of the estimators. How can I estimate the Covariance Matrix when I have the theoretical one ? Can i just plug in the estimators of my coeficients ?

[startz]

Hi all,

I have a question how can I estimate an ARMA(2,1) model in Eviews ?

Further I have calculated the asymptotic distribution of the estimators. How can I estimate the Covariance Matrix when I have the theoretical one ? Can i just plug in the estimators of my coeficients ?

Code: Select all

ls y ar(1) ar(2) ma(1)

Gives a quasi-maximum likelihood estimate and also gives the estimated covariance matrix.

[hxh]

but actually i want to have the maximum liklehood estimators and their standards errors, this is Least sqaures

[startz]

With the exception of treatment of the first two observations, least squares is MLE.

[hxh]

Thats what I should answer

A researcher fits an ARMA(2,1) model for the mean adjusted growth rate yt = yt-\bar y
(without deterministic terms). Use the result on the asymptotic distribution of ML
estimators in stationary ARMA models and derive an expression for the asymptotic
covariance matrix SIGMA/T of the estimated parameter vector beta . It is
sufficient to give the result in form of the matrix S such that SIGMA=S^-1.
Estimate the covariance accordingly and compare the implied asymptotic standard errors with those
reported by EViews.


I dont really understand what I have to do, because for sure I can get this Covariance MAtrix by using OLS, but how can I compare this to my theoretical result?

[startz]

Thats what I should answer

A researcher fits an ARMA(2,1) model for the mean adjusted growth rate yt = yt-\bar y
(without deterministic terms). Use the result on the asymptotic distribution of ML
estimators in stationary ARMA models and derive an expression for the asymptotic
covariance matrix SIGMA/T of the estimated parameter vector beta . It is
sufficient to give the result in form of the matrix S such that SIGMA=S^-1.
Estimate the covariance accordingly and compare the implied asymptotic standard errors with those
reported by EViews.


I dont really understand what I have to do, because for sure I can get this Covariance MAtrix by using OLS, but how can I compare this to my theoretical result?

I imagine that you're being asked to plug the estimated parameters into the expression for the asymptotic covariance matrix and compare the diagonal elements to the standard errors reported by EViews. But you might do better by asking for clarification from whoever assigned the homework.

[hxh]

yeah i also think i should do that