The poster “The Magic of Data Analysis” shows four different tools within MAXQDA for analyzing documents: Code Frequency, Code Matrix-Browser, Code Relation Browser and the Document Comparison chart. In this article we will have a look into the actual .mx-projectfile and understand how these results have been created.
Back in 2017 I have created my first poster featuring some functionalities of MAXQDA and how I used them during my master thesis (Müller 2017) on witchcraft documents from Rostock 1584 and Hainburg 1617/18. Back then I was euphoric about my results and prested them at the MQIC2017. Today I think that with my thesis I toppled head over into the Positivist Pitfall (just describing what is there instead of interpreting it) but the results in my eyes are largely still valid and they demonstrate well how QDA software can be used.
Today the poster is widely available on several places on the web (Müller 2017a). However, what I have never documented yet is how I created these results in MAXQDA.
My data was split into two groups of document: 18 coming from Rostock 1584 and 19 from Hainburg 1617/18. They were especialy well suited for QDA as these historical court records had a very clear and repetetive structure. These documents are made up by numbered paragraphs that summerize the “confession” of the accused person. In my analysis I always attributed codes to the entire paragraph so that I could describe their numerical quantity based on their pre-existing internal structre.
As codes I chose five (deductive) codes coming from literature of witchcraft research: pact with the devil, sexual intercourse with the devil, witches’ flight, witches’ meeting and harmful magic. For the last category I created inductive subcategories to differentiate this very general category.
For comparnig the code frequencies of the two document groups (Hainburg and Rostock). I chose the “Compare Groups > quantitative” function and selected the document groups “Hainburg” and “Rostock” as well as the five codes including their subcodes. The result looked like this:
For the poster I polished it up a little and especially highlighted the differences in the subcodes of harmful magic as it was the easiest result to understand.
Code Matrix Browser
The Code Matrix Browser is strongly similar to the Code Frequency but here I visualized individual documents instead of groups. To create this result I selected (activated) the 18 documents from Rostock and chose to illustrate the code “harmful magic” with its subcategories. I started the Code Matrix Browser and set the visualisation so that the size of the symbol refers to the row. the The result looks like this:
Here, I highlighted the increase of damage against animals i Documents R09-R12 as the most easy to understand result.
Code Relation Browser
The Code Relation Browser visualizes where codes are used on the same textsegment. In my setup I decided that the five main codes are allowed to overlap (this was analytically interesting) but the subcodes of harmful magic were not allowed to overlap (this would invalidate their quantity). Unfortunately these results were hard to understand or interpret for an outsider, therefore I chose the most obvious and intuitive result. I activated the documents for Rostock and the five main categories (ignoring overlap with the subcodes). In the settings I choose “choose top level codes” and selected only the first four codes to avoid unwanted “noise”. I ran the Code Relation Browser and the result looked like this.
I highlighted the overlap of witches’ flight and witches’ meeting (as something easy to expect) and the overlap of harmful magic and sexual intercourse with the devil as a more surprising but still easily understandable result.
Document comparison chart
For the Document comparison chart I asigned one colour to each top-level code. Because the documents in Hainburg followed a very clear structure also in relation to their content, their were perfectly suited to demonstrate that feature. I activated all the documents from Hainburg and all the codes and started the Document Comparison chart. The result looked like this:
As the image was both illustrative and meaningful I just had to explain the colour code and demonstrated the overall appearing structure of the documents.
As we can see the results I chose have been highly selecitve. I always chose that result which visualized best and was easy to understand for an audiecne not familiar with my topic. This is of course necessary when presenting results to a wider audiecne. However, the images are genuine and hardly needed any external changes. They can be easily replicated within MAXQDA and served me well not just in visualizing my results but also in analyzing my data. As all there diagrams are interactive within maxqda they can be used to trac down outliners or look at a group of specifically intersting codes. For example a click on the node at “sex” and “harmful magic” immediatly calls up all segments where these two codes overlap.