Scoring cases with missing values

[4Walk]

I am trying to understand how SAS Enterprise Miner scores cases that have a missing value in one of the predictors.  Here is a small regression example.  

The SAS EM estimated regression equation is: 50.8706 - 2.5578X1 + 0.0978X2

For the case: X1 = 8 and X2 = . SAS EM gives a predicted value of 41.01.  I am not able to understand how this number is calculated.  If the missing value of X2 is set to zero, y-hat must be 30.41.  To get the value predicted by SAS, the predictor X2 must be 108.445.  This is way different from the average of X2 which is 93.649.

This is a very simple diagram, data source, partition and regression node, that's it, no impute node.  I thought the cases with missing values will be omitted by regression node, but apparently not.  I am not understanding something.

Thank you .....

Thank you.

[WendyCzika · Accepted solution]

When any of the predictors is missing, the predictions for the target values are the prior probabilities of the target (if class) or the average target value (if interval).  The priors and average are calculated across the obs. with no missing predictors.