During the past 3 weeks of testing out our collaborative NVivo coding scheme, we have not only solidified our codes and their scope, but have had a chance to compare our coding styles. The big disparity at this point is the size of the coding chunks that are selected by each coder. Some are more inclusive of surrounding context, while others are tighter in their coding style.

We are pushing some of the task onto NVivo’s shoulders. The software allows us to code tightly (i.e. only the parts of a transcript directly relevant to the code), then easily “open up” to see more context around the coded selections.

As we move forward we will be creating new codes beyond our core set, but they will be added on a per-paper basis. The clean, consolidated, coded dataset will serve as a launch point for writing a paper. We’ll add new codes  and share those among the authors of that paper to keep the central repository clean.

Now we just need to code like the wind! It’s tedious and time-consuming, but the activity itself expands your understanding of the interviews.

Here’s the post from our October 2 call, delayed by the usual delay-o-rama:

Today’s conversation about coding was meant as a 10 minute “touching base” about how our coding is going and whether the codes are working for the interviews we have done. It turned into a lengthy discussion about the architecture of our codes and the use of codes to house disjunctive categories.We had some debate about intercoder reliability. On the one hand, such reliability seems difficult/impossible to achieve in our project and that codes mainly serve as signals for others on our team. On the other hand, we were concerned with having a large disconnect between the way each individual is using the codes, such as large differences in the number of codes used in one interview. We will revisit this in a couple of months and see what coding style differences are like after we have done more interviews.

We talked about recording our own conversation, with some scepticism about whether we would ever actually go back to the recordings. We will try it and see.

Our coding scheme is co-evolving with our understanding of it. This alone would make it difficult to share our data with others (an issue also facing some of our subjects).