|
|
I'm running a time series model (TSMODEL) on several groups of data. As mentioned in another post, if I have the procedure autodetect outliers, I sometimes get crazy predictions.
By looking at the R-squared, you can see when the model is poor (of course). In one case, I get an R-squared of -3996 and in another I get -600.
What I would like is to be able to somehow flag those cases...maybe somehow save the R-squared? And then I could re-run the model on those cases and omit the outlier detection.
If nothing else, I guess I could compute my own R-squared and save it. But I was wondering if there was anything that might be a little more simple.
|
Locked
