The idea of creating/refining algorithms to perform some version of qualitative coding is not new or novel, although the practice does not seem yet to be widespread enough for people to consider it routine. I suspect this is due in part to the fact that many people who do research have been trained to be project managers, rather than technicians, and many technicians focus on methods for their own sake, so are not in the habit of thinking through research questions and designs to the extent needed to publish or fund a project.

And, as an aside, I like to think of myself as a little more of a generalist or maybe "working supervisor" is more apt. I also think methods training is more important in a graduate degree than subject matter training. I suspect many of my colleagues differ in their views.

But, to return to my subject matter of interest for this post - I want to begin by acknowledging that  many qualitative researchers have described, in writing or in conversation, the reasons why a program cannot code or analyze data as well as a human. A recent experience with (human) coding of a biggish data set reminded me of several of these.

I have been struggling for a few days to copy some text information from a moderately large Excel spreadsheet  - it has 1200 + rows and the columns go from A to BO - using Word's mail merge function.

I first discovered the joy of mail merge when I was creating vignettes from an Excel data frame I made using R. I used R to randomly order category levels and printed the options to Excel. Then I added some stem wording in Word, and was able to end up with what looked like a paragraph of text but was really a fairly complex multiple regression equation with a lot of categorical predictor variables.

Once I started to do mixed methods secondary analysis with certain datasets from the Centers for Disease Control and Prevention, I began copying and pasting narrative data from Excel into Word to improve my ability to read and apply qualitative coding processes. Then one day - and I do not know why this idea suddenly came into my mind -  I realized that I could also use mail merge to print the information from the cells I wanted, into a series of Word documents. I did this over the summer in bundles of 100 with no real problems. I was working in 100s because I wanted to give data to my student research assistant incrementally. However, earlier this week when I tried to print the content from just 3 columns of all 1200 + rows, I immediately got error messages, and only a portion of the cases actually printed (not in order, unfortunately). I also got a lot of this: