PREDICTION ACCURACY
Posted: Tue Dec 13, 2011 5:22 am
Hi everyone. This is what I have to do for my job:
I've got (say) two series inflation(inf) and output gap according to different definitions (og1, og2, og3 etc...).
To judge the prediction occuracy of each measure of OG upon INF, i want to estimate
INF=c(1) + c(2)OG
with a rolling regression of this kind:
-200 initial observation
-cut the last 20, I got 180 obs left (from 1 to 180)
- use 1 to 180 to estimate the parameters;
- use parameter estimates for 1-step forecast (181° to be estimated)
- calculate between forecast and actual value and save;
-cut the last 19, I got 180 obs left (from 2 to 181)
- use 2 to 181 to estimate the parameters;
- use parameter estimates for 1-step forecast (182° to be estimated)
- calculate between forecast and actual value and save;
and so on, as long as I can still cut the sample.
Then I need a RMSE (root mean square error) computed on all the residuals saved!
CAN ANYONE PLEASE HELP ME WITH THIS? I NEED IT AS QUICK AS POSSIBLE!
I've got (say) two series inflation(inf) and output gap according to different definitions (og1, og2, og3 etc...).
To judge the prediction occuracy of each measure of OG upon INF, i want to estimate
INF=c(1) + c(2)OG
with a rolling regression of this kind:
-200 initial observation
-cut the last 20, I got 180 obs left (from 1 to 180)
- use 1 to 180 to estimate the parameters;
- use parameter estimates for 1-step forecast (181° to be estimated)
- calculate between forecast and actual value and save;
-cut the last 19, I got 180 obs left (from 2 to 181)
- use 2 to 181 to estimate the parameters;
- use parameter estimates for 1-step forecast (182° to be estimated)
- calculate between forecast and actual value and save;
and so on, as long as I can still cut the sample.
Then I need a RMSE (root mean square error) computed on all the residuals saved!
CAN ANYONE PLEASE HELP ME WITH THIS? I NEED IT AS QUICK AS POSSIBLE!