Nvivo calculates inter-rater reliability based on how much agreement coders have on a per character basis. That is, how much do they agree to highlight certain sections of the sentence. This is NOT the unit of analysis that I need. I need to conduct a coding comparison based on the codes used per QUESTION or per block/paragraph of text. Has anyone found a work around that allows the unit of analysis to be based on question or block of text INSTEAD of per character?