I'm trying to do an expanded uncertainty calculation for my Agilent 3458 Digital Multimeter. I noticed when looking at the specifications that there are many errors including RMS noise, linearity, and a temperature coefficient as well as other sources of error. When doing an overall expanded uncertainty calculation, do I include all these errors for an accurate representation of measurement uncertainty? If so, do I combine all these errors in a linear combination, or do a root-sum-squares analysis treating them each as a separate error source? I ask this mainly because in one of the examples on the first page of Appendix A the RMS error is added directly to the accuracy specification.