It's vital to have quality checks in any calculation step along the way, so make sure to check for bad values before outputting values to attributes. You question is, however, rather short so please add more information (and a screenshot if possible) so we can answer this more in context.
Agree - having data quality checks embedded in the AF analysis would be a good approach. You may also want to have an analysis that checks for data quality issues for any attribute or variable used in the analysis so these can be flagged for further review.
