Announcement

Collapse
No announcement yet.
X
  • Filter
  • Time
  • Show
Clear All
new posts

  • Adrian Cernescu
    started a topic How to use ACF and PACF the right way ?

    How to use ACF and PACF the right way ?

    Hi, I have looked online and seen different versions of how to use ACF and PACF. I am running the following regression:
    Code:
    reg CSAD X1 X2 X3
    if I want to find visually the number of lags I should use for CSAD in my regression to alleviate autocorrelation which method is the appropiate one?

    method 1:
    Code:
    ac CSAD
    pac CSAD
    And from here look and decide the number of lags

    method 2:
    Code:
    reg CSAD X1 X2 X3
    predict residual, r
    ac residual
    pac residual
    And from here decide how many lags of CSAD i should use.

    my intuition are that the two are the same, and running method one will simply show with what lags of itself CSAD is correlated, whereas running method 2 will show to how many lags will the error term be correlated with past values, meaning that the error term includes some variables correlated with CSAD, more specifically its lags that are given in method 2.

    could someone please tell me if I am right or which method I should use.

    thank you

  • Adrian Cernescu
    replied
    Originally posted by Joro Kolev View Post
    You would use Method 1 if this was a pure ARIMA model, without exogenous regressors X1 X2 X3.

    As you have exogenous regressors in your model, you are interested in the ARMA properties of the errors in the model after you control for the exogenous regressors. This is why Method 2 should be used here.


    Thank you for your response, this makes a lot more sense for me now, I appreciate the time taken to answer my question.
    All the best,
    Adrian.

    Leave a comment:


  • Joro Kolev
    replied
    You would use Method 1 if this was a pure ARIMA model, without exogenous regressors X1 X2 X3.

    As you have exogenous regressors in your model, you are interested in the ARMA properties of the errors in the model after you control for the exogenous regressors. This is why Method 2 should be used here.

    Originally posted by Adrian Cernescu View Post

    Hi Joro Kolev, could you please provide me with an intuition behind why should I use method 2, I would much appreciate some insight.

    thank you.

    Leave a comment:


  • Adrian Cernescu
    replied
    Originally posted by Joro Kolev View Post
    You should use Method 2.
    Hi Joro Kolev, could you please provide me with an intuition behind why should I use method 2, I would much appreciate some insight.

    thank you.

    Leave a comment:


  • Joro Kolev
    replied
    You should use Method 2.

    Leave a comment:

Working...
X