NeuroShell 2 gives you several methods of dealing with missing data in the Training Criteria module. Click on the appropriate box to either replace the missing value with a 0, the variable's minimum or maximum value, or an average of the minimum and maximum values. Choose whatever is considered to be the most likely value for the variable.
If you expect missing values in your data, we recommend using an average of the minimum and maximum values, which is probably the best estimate of the value. The neural network works better if you make an intelligent guess of the value rather than leaving it blank.
If you're not expecting missing values, check Error Condition (the default setting) to alert you to missing values in your data.