turn on suggestions

Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

Showing results for

Find a Community

- Home
- /
- Analytics
- /
- Stat Procs
- /
- ARIMA - prewhitening procedure for studying causal...

Topic Options

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Highlight
- Email to a Friend
- Report Inappropriate Content

08-13-2009 12:52 AM

Hi I wonder if anyone could check my coding of the following models. The models are used to study the causal relationships between two time series, x and y. Only one of the two time series, x, follows an MA(1) process and needs prewhitening.

The first model shown below regresses y on four input series, the concurrent value and 3 lags of the prewhitened series x:

yt = a-1 + a0 xt + a1 xt-1 + a2 xt-2 + a3 xt-3 + et

I follow the prewhitening example found in the SAS documentation to build the following lines:

proc arima data=in;

identify var=x;

estimate q=1;

identify var=y crosscorr=x;

estimate input=(x x(1) x(2) x(3))

run;

quit;

My concern is that in the SAS documentation, it suggests that the above procedures will filter both x and y by the MA(1) model. As I need to regress the unfiltered y series on the prewhitened x series (concurrent plus 3 lags), I am not sure if this requirement corresponds to the 2nd estimate statement.

To complete the study on causal relationship, I also need to swap the input and response series as follows:

xt = b-1 + b0 yt + b1 yt-1 + b2 yt-2 + b3 yt-3 + ut

I wonder if the following codes are correct:

proc arima data=in;

identify var=x;

estimate q=1;

identify var=x crosscorr=y;

estimate input=(y y(1) y(2) y(3))

run;

quit;

Thank you

The first model shown below regresses y on four input series, the concurrent value and 3 lags of the prewhitened series x:

yt = a-1 + a0 xt + a1 xt-1 + a2 xt-2 + a3 xt-3 + et

I follow the prewhitening example found in the SAS documentation to build the following lines:

proc arima data=in;

identify var=x;

estimate q=1;

identify var=y crosscorr=x;

estimate input=(x x(1) x(2) x(3))

run;

quit;

My concern is that in the SAS documentation, it suggests that the above procedures will filter both x and y by the MA(1) model. As I need to regress the unfiltered y series on the prewhitened x series (concurrent plus 3 lags), I am not sure if this requirement corresponds to the 2nd estimate statement.

To complete the study on causal relationship, I also need to swap the input and response series as follows:

xt = b-1 + b0 yt + b1 yt-1 + b2 yt-2 + b3 yt-3 + ut

I wonder if the following codes are correct:

proc arima data=in;

identify var=x;

estimate q=1;

identify var=x crosscorr=y;

estimate input=(y y(1) y(2) y(3))

run;

quit;

Thank you