When fitting a model, PROC GAM excludes any observation with missing values for an explanatory variable, offset variable, or dependent variable. However, if only the response is missing, predicted values can be computed and output to a data set by using the OUTPUT or SCORE statement.