Having changed the format has proved the difference in rounding is in the formats. As a pitty it is not well documented how rounding is done behind scenes. data_null_ has given another example how rounding effects can give odd results. The - sign is a nice example. To be added: We are into the digital ages no not digital numbers but binary numbers. https://support.sas.com/techsup/technote/ts654.pdf Your decimal fractions are not possible to be represented exactly in their binary counterparts. That will introduce some of precision differences as data_null_ coded manual. Having the summing done as integers on dollarcents would have avoided fractions. The format Best. (that default) is said to adjust to most applicable resolution. That is all what is said about that. SAS(R) 9.4 Formats and Informats: Reference More interesting is the SAS(R) 9.4 System Options: Reference, Third Edition decimalconv= system option. It refers to: IEEE Standard for Floating-Point Arithmetic 754-2008. SAS did not conform to that but can now.
... View more
