Crosstabs or matrix coding on multiple (3) distinct nodes? Is this possible?

[kmllr]

Hello,

I'm attempting to figure out whether there's a way to run a cross-tab or matrix query on three distinct nodes (not attributes) simultaneously. For instance, I'm attempting to track the relationship between node X and time (where time is also a node, coded by year) by node Y. I've attempted to treat each node X year and time as a set (both using the nodes themselves and using a coding query of the two turned into a set), and then running a matrix query, but, unfortunately, due to the structure of my data, this is not working. Essentially, as my data is reall asymmetric, I've not been able to use node attributes to sort what might be understood as a "case" (I'm not using interviews, I'm doing discourse analysis over a huge variety of sources), and, unfortunately, multiple instances of the same node (or variations on a node) are coded within the same source. For instance, I have a source that is coded uniformly for node X, but contains three separate years. As a result, when I attempt to run the matrix query utilizing either of the sets I've created, it returns the results for node X and all three years, rather than the one I'm interested in, by node Y. This is pretty frustrating, but as I've spend months on the coding, I'm not able to attempt to restructure the data. Also, I'm interested in this type of three node cross-tab overall, not just for something that structures data, like year. I've realized that I could run a bunch of separate compound queries and then chart the results in Excel, but I'm hoping there might be an easier way to cope with this within the program?

 

Any light that could be shed on this issue, I would greaty appreciate. Similarly, if the query isn't clear, please let me know, and I shall attempt to provide additional information.

Thanks so very much and best wishes,

Kate

[QDATRAINING Admin]

Hi Kate,

I think what you need is what I refer to in workshops as a 'compound matrix'. But before I send you instructions, I need to be sure I'm on the right track.  I'm gong to send you a link, password and instructions to upload your project file to our secure server so I can see these nodes and ensure my solution works on your particular dataset. I'll send these privately by e-mail as I do not want to publish these instructions. However, if the solution works, I'll publish the instructions as a reply on the forum so others who may be having a similar problem can benefit from the response .

Kind regards,

 

[QDATRAINING Admin]

We did not receive any NVivo project file from this posting so could not check the veracity of a compound matrix in this case. For anyone reading this post, a compound matrix is where you run an NVivo matrix against two sets of nodes and convert the results into a new set of nodes. You then add your new nodes to a new matrix to further intersect them with other nodes or demographics/profiling or questionnaire data. 

For example, if I had attitudinal, behavioral or similar qualitative meanings coded to a nodes or indeed several nodes such as types of risk for example. And if I wanted to understand the degree to which risks were gender and age related in my study population. My fist matrix would sub-divide my risk types into sub-nodes demarcated along gender lines. Now I can see if women in my group share or take different types of risks to men. I convert my matrix to nodes and put the gender split nodes into an age range matrix. Now I can see if younger or older men, or indeed women,  take similar or different risks or if age mitigates risk based on how much data comes from younger or older men and/or women. 

Hope this is helpful to anyone considering this type of analysis

[Author]

Posts