My relationship with NVivo 9 has become combative; it’s an outright, unsanctioned fight under MQCA rules (mixed qualitative coding arts). The report I needed from NVivo 9 was a list of all coding references with each having a list of the codes/nodes to which it was assigned. I’ve already spent one, previous post ranting about the fact that NVivo 9 can not perform that task: one that seems to be a basic necessity of qualitative research and analysis. It took more than a few hours of tinkering and learning, but I’ve managed to learn how to accomplish a reasonable facsimile of that report. I’ll try to describe what I’ve done in enough detail so that it may be helpful to others perhaps struggling with or learning NVivo 9 as I am. Read on if interested; otherwise, return your browser to its regularly scheduled viewing activity.
Several things I had already done in setting up and coding my data made this report possible:
- I imported a single source: an Excel file. Column 1 was participant ID, Column 2 was pretest survey response, Column 3 was posttest survey response.
- I used the autocode feature to create a node for each row/participant. Those nodes were labeled sequentially from 10001 to 10370 and grouped together under a parent node labeled “Participants.”
- I created a node classification called “Person” which I applied to all of the participant nodes. The classification contained demographic attributes and quantitative attributes based on likert scale survey items.
- When I coded, I highlighted the entire pretest response/cell for a participant and then used “Code at Existing Node” or “Code at New Node.” In addition to the many different possible nodes/codes specific to the text submitted by the participants (the nature and content of the response), I also coded the response to the corresponding participant node created as described above.
To generate the report, I used a Matrix Coding Query with the following settings:
Matrix Coding Criteria
- Rows Tab > Selected Items > Select and add all participant nodes as rows.
- Columns Tab > Selected Items > Select and add all codes/nodes as columns.
- I didn’t do anything on the Matrix Tab.
- Changed Results Option to “Create Results as New Matrix”
The result is a VERY large table: 370 rows and 100 columns; the 370 rows being participants and the columns being the codes/nodes I created as I read and coded the survey responses. By default, the number shown in cells – at the intersection of participant node and coded node – is the number of sources; right-clicking on a cell and selecting Matrix Cell Content > Coding References will display how many coding references are coded at that intersection.
Two benefits of this report. First, it can be exported for further analysis. I’ll probably do that and use VLOOKUP in Excel to connect the participant to the demographic data and work with Pivot Tables to view and rotate the data in several different ways needed for my analysis. Second, the matrix report has drill down capabilities. Double-clicking on a cell will dynamically generate a report listing the coding references that substantiate the intersection of participant node and coding node.
And, I’m not getting out of this post without one complaint. Instead of having rows as participants and columns as coded nodes, I originally wanted to do the rows as as coded nodes and columns being 3 different attribute criteria queries. For example, I have attribute data that is the participants average self-assessed rating of computer literacy skill; those range from 0.2 to 5.0. I wanted to have three columns for average self-assessment score: (a) < 2.5, (b) 2.5 =< AND < 3.5, and (c) >=3.5. The intersections/cells would have revealed how many users self-assessing below average gave a response coded to each node – revealing themes within the responses of users at different skill levels. So when entering selection criteria, why or how in the world is it not possible to select criteria that is AND or OR of an attribute? I can select < 2.5 and >= 3.5, but I can not create a column that does 2.5 =< AND < 3.5. Goofy.