Predicted values can only be computed when all predictor values are in the range the data used to fit the model in PROC GAM. If the value of a predictor variable in the SCORE DATA= data set is outside the range of that predictor variable in the PROC GAM DATA= data set, then predicted values for that variable and for the response will be missing in the SCORE OUT= data set. Note that PROC GAM can also produce predicted values for new observations that are included in the PROC GAM DATA= data set with missing response. However, these predictions will also be missing when a predictor value is outside the range of the predictor in the observations used to fit the model (those observations with nonmissing response).