Forecasting using SAS Forecast Server, SAS/ETS, and more

Collinearity diagnostic for ARIMA intervention model ?

Reply
N/A
Posts: 0

Collinearity diagnostic for ARIMA intervention model ?

Dear all:
I have a question about how to do collinearity diagnostic when I perform ARIMA intervention analysis.
My research is to investigate the impact of National Healthcare policies on drug utilization. Five policies were implemented during the study period. Because of the existence of unit root, so I select ARIMA intervention model. I specified two kinds of interventions (step function and ramp function) for each policy to model the change in level and change in trend. Therefore, a total of 10 regressors were specified. However, when I run intervention model, it shows that some parameter estimates were highly correlated and collineraity must be considered.
My question is:
1. How to write SAS syntax to check "collinearity" ?
Because I search for "Help and Documentation" and I only can see the following syntax:
proc reg data = "data_name";
model y = x1 x2 .../tol vif collin;
run;
However, this syntax seems for linear regression and not for ARIMA intervention model.
2. If I use the results of ARIMA procedure to check collinearity, only "correlation of parameter estimates" can be used. However, I don't know how high of the correlation should be considered the existence of collinearity.
I am looking forward to getting help from you.
Thank you!
-------------------------------------------
teresa, Huang, PhD student
Ask a Question
Discussion stats
  • 0 replies
  • 441 views
  • 0 likes
  • 1 in conversation