Here are a couple of documents that might be helpful for categorical data:

https://www3.nd.edu/~rwilliam/stats1/x51.pdf

http://web.cs.dal.ca/~anwar/ds/Lec3.pdf

and

http://ocw.jhsph.edu/courses/hsre/pdfs/hsre_lect7_weiner.pdf

Which also lists some references.

Take a look at this paper:

Lombard, M., Snyder‐Duch, J., & Bracken, C. C. (2002). Content analysis in mass communication: Assessment and reporting of intercoder reliability. Human communication research, 28(4), 587-604.

Negative scores, usually, mean that you have little variability in codes AND there are 1 or 2 disagreements by one coder.

Also, seven categories might be tough to code. Have you considered collapsing them?

Have you looked at Landis, J., & Koch, G. (1977). The Measurement of Observer Agreement for Categorical Data. Biometrics, 33(1), 159-174. ? It is a pretty standard reference for this sort of thing.  I hope you have a constant number of raters; in that case, you can use Stata to derive a kappa, as in  http://www.reed.edu/psychology/stata/gs/tutorials/multiple.html 

which follows the method outlined in Landis & Koch.