s**********l 发帖数: 395 | 1 Hi, All
I have a might-be studpid question about adjusted R-square usage in SAS :
The case is below:
We use a sample data (Jan.2003-Dec.2010)to built a model:
Model: General linear regression model
Target variable:Y
Input variable: x1, x2,x3,....x8
Predicted variable of Y: Y_hat
And then we apply this model to out_of_sample (Jan.2011-Dec.2011)data to
generate predicted value for Y1, called Y_hat1.
Now we want to test whether the model fit the out_of-sample data well, we
need to compare how different the predicted value y_hat1 and y1. My
colleague test in this way using SAS:
Proc Reg data=a ADJRSQ MSE;
model y_hat1=y;
run;
Her judgement is that if Adjusted R square is higher, then the model fit the
out_of sample data.
I told her that she was wrong, but she insisted that she was right and told
me that from business point, we can use in this way.
Is anyone who see this usage of Adjusted_Rquare before? Thank you. |
|