Ruppenthal, Marc wrote:

[...error propagation...]

Hi Marc,

regarding your question: you have to use the standard deviations (s.d.) 
in the formula.

But maybe a word of caution at this point:
Do not overstrain the statistical calculation of errors. Of course, 
correct mathematics are essential for a proper error calculation and 
this undoubtedly necessary. However, do not try to minimize your errors 
using some statistical voodoo. Always keep in mind what a statistical 
parameter like SD or SE tells you. As Hilary wrote:

 > I think that severe statistical treatment may give false confidence 
in > results.

regards,
Robert