Our qualitative archive provides new analyses of the themes discussed across locations and years. Some of the main analysis tools we’ve relied on are frequency and cooccurrence data, topic modeling, dendrograms, and link analysis. In this post, we’ll share how each of those analyses functions and some of the most interesting findings within the archive found through each method. 

WordStat can either show the frequencies of the most common words and phrases (excluding those which come up incidentally) or can show the frequencies of words and phrases specified in a list created by the researcher called the categorization dictionary. Frequency refers to the number of times a given word or phrase appears in the archive. Categorization dictionaries allow searches for lists of words and phrases created based on the project. For example, a dictionary seeking to understand how people with disabilities relate to employment might include words like disability, employment, job, and career. Frequency is useful for knowing how often participants are discussing a topic, the distribution of the topic across cases, and the percentage of participants who discussed the topic.  

The cooccurrence data draws on the frequency data to illustrate where in the archive words and phrases are discussed together. This analysis can be done at the sentence, paragraph or document level, giving researchers a multitude of ways to understand connections between topics. This cooccurrence analysis can be as simple as comparing the number of times topics come up together within the archive or can culminate in a visual representation like a dendrogram or link analysis.  

Based on the cooccurrence data and the strength of the association, WordStat groups words and phrases together. Seeing how words relate to each other in a dendrogram can illustrate how participants connect ideas and experiences. Link analysis is another way to visualize these relationships, depicting each word as its own bubble and illustrating the strength of its individual ties to other topics in the cluster. Topic modeling similarly groups words together based on cooccurrence and assigns them a topic that describes the most common use of the terms. These analyses provide big picture representations of the archive and insight into the themes that participants have discussed the most frequently over the years. 

By Anika Lanser, Research Assistant at the Center for Social Development and Education