This is the reply from support:
QlikView uses IEEE 754 64-bit double-precision binary floating-point format to store all numbers. This means that 64 bits are used to store a
number: One bit for the sign, 11 bits for the exponent and 52 bits for the number itself. A 52 bit number corresponds to approximately 15 significant
decimal digits. This means that QlikView can store integers up to 14 digits while preserving the exactness of the number. However, integers with more than 14 digits will be rounded to 14 significant digits.
For non-integer numbers, the situation is slightly worse. Some fractional numbers can be stored in an exact form, e.g. 0.5 and 0.25, since
they have an exact binary counterpart. But others cannot, e.g. 0.1 and 0.2, since these do not have an exact binary counterpart. As a consequence, you very often have a “rounding error” in the 14th significant digit for non-integers...
These are not errors in QlikView. And they are not errors in IEEE 754 either. Rather, they represent errors in the expectation of binary floating point
numbers. It's simply that some values cannot be exactly represented as binary numbers, so you get rounding errors. There's no way around it...except for first converting the floating point number to an integer (by multiplication), leaving no decimal part, and thereafter rounding the integer accordingly and finally converting the integer back to a float (by division)…
Normally this kind of "rounding errors" are small and usually insignificant and should in most cases not influence the result of the analysis.
