1
$\begingroup$

I am new to the subject and trying to learn and equip well into the topic. I got a problem to solve and it only contains the model equation - {Generic AR(N) model} modelled using the equation: v[k] + d1v[k-1]+.......dNv[k-N] = e[k], where e[k] is the white noise sequence. How to estimate the coefficients for N = 3?

Any pointers on this topic will help or steps to validate the solution.

If I have some model parameters , d1,d2,d3, how to verify them for the model of N= 3

Improve this question
$\endgroup$
3
  • $\begingroup$ You can set it up as a least squares problem (you have probably heard of OLS) with v[k] as the dependent variable and v[k-1], ..., v[k-N] as regressors. This is usually covered in time series textbooks, take a look at some of them. $\endgroup$
    – Richard Hardy
    Commented Nov 19 at 8:31
  • $\begingroup$ Thanks @RichardHardy. Is there a technique called as Iterative Least Squares Method? Can I perform these calculations by-hand instead of using any software tool? (Mainly to understand and get an intuitive understanding of the approach. ) Any standard textbooks that you can advise to look into, containing numerical-problems , examples to try-out $\endgroup$
    – eashwar natarajan
    Commented Nov 19 at 23:32
  • $\begingroup$ The technique exists but is not needed for estimating AR models, because simple least squares works just fine. I do not have a particular textbook to recommend, but here is a general list: stats.stackexchange.com/questions/20514 $\endgroup$
    – Richard Hardy
    Commented Nov 20 at 8:02

0

Reset to default

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.