I am studying stock price.
I first define that the ln(price) has an obvious unit root of 1 lag.
I was wondering if someone could explain me how to proceed to construct a correct ARMA model...?
Thanks is advance.
Add a comment
|
1 Answer
15
-
$\begingroup$ Thanks.... My doubt is about using the difference of the lnprice as dependent variable, > D(lnprice) c AR(1) .... or the normal lnprice and how do I know how many lag to use in the MA or AR. This is the first difference correlogram I dont see any particular behavior at a first sight. Do I need to look at the lnprice? postimage.org/image/89yunf647 $\endgroup$ Commented Mar 20, 2013 at 0:07
-
$\begingroup$ Yes you need to look at the lnprice not the difference. Please accept the answer if you are convinced. $\endgroup$– MetricsCommented Mar 20, 2013 at 0:13
-
$\begingroup$ so in this case you will run the arma model with lnprice as dependent variable? $\endgroup$ Commented Mar 20, 2013 at 0:40
-
$\begingroup$ You need to look into autocorrelation (ACF) and partial autocorrelation function (PACF)to determine whether it is AR(1) (or ARMA(1,0)) or MA(1) (ARMA(0,1)) or ARMA(1,1,) or some higher order times. If you are not sure,do update your question with ACF and PACF, so that we can help you. What is your objective? $\endgroup$– MetricsCommented Mar 20, 2013 at 0:47
-
$\begingroup$ I need to define the arma, but I dont understand if I need to take the first difference or use the non-stationary lnprice as dependent variable. I thought you needed a stationary series for arma. I a little bit confuse on the step to follow. $\endgroup$ Commented Mar 20, 2013 at 10:19