09-15-2016 11:03 AM
Can anyone tell me why EG compare datasets will show a very small difference for an exact match and if there is a setting that might be doing this that I should change? I have attached a snapshot of the results. The Base plot variable is a calculated field BUT it is only created by adding .01 NOT multiplication.
09-15-2016 11:10 AM
Seems a Machine-Precision probleme. It's related with how number are saved in base 2 in a compute, for example 0.1 could not be represented without error in a compute.
It's happen in all compute with all numbers, a way to solve it could be a round(var,0.0000000000001). Since you are working with not many decimals, this should not affect your calcs and solve the problems with EG.
09-15-2016 12:00 PM
The Proc Compare option FUZZ allows you to tune how close values may be to consider equal and not make into the report.
Try adding (or selecting the option) to set Fuzz=1e-10. That would consider values within 0.0000000001 to be equal. Since 1e-10 is larger than your shown e-14 range then likely most or your issue would be resolved.