Errors (if present) are amended using the local linear approximation:
sigma[f(x)] = sigma[x] * df/dx .
If input is primitive user will be asked to give error estimate. If none desired, type "!" and not RETURN which would give error of 0.0 If data is negative and LOG10 or LOG is used, then data is set to 1.E-38 and error such that when added to data it will have value plus 1 sigma. If data is negative and SQRT is used, then data is set to zero and error modified.
If EXP or 10** are used and data is such that Arithmetic Overflow would occur, data values and errors are replaced to indicate that this is the case. The data label (if present) is amended to reflect the operation performed, and all other components are unchanged.
Subtopics: