How to Choose a Visualization

What is Coding In the Context of Qualitative Data Analysis?

"In social research, coding simply means making and applying categories to your data. You can use your code to develop comparisons and connect your data to relevant theoretical concepts you've located in your literature review. Taken together, these codes, comparisons, and concepts help you build explanations that address your research questions." [source]

Coding leads to more meaningful insights about your data, but it can take a lot of time and skill.  These are visualizations that are possible without coding your data, although coding is recommended. The visualization descriptions also discuss their shortcomings. 

Word Count Bar Chart

Example:

Word count bar charts reduce some of the bias created by word clouds, although they are less aesthetically appealing. They also share some of the same issues:

If you do not do the additional work to clean the data

1. They do not capture synonyms, words that mean the same thing (i.e. fast, quick, speedy)
2. They do not lemmatize words, group variations of the same words (i.e. eating, ate, eat, eats)\
3. They may not remove stop words, depending on the software you use (i.e. and, on, or, the, etc.)

Regardless of cleaning the data

1. They do not capture themes
2. Related to #1, they do not show word pairs or groups which can provide important context (i.e. not cheap, costs too much, make more affordable)
3. They lack context - who said what, when, how, where, what? What was the connotation? How many times did they same person say the same thing?

Word Clouds

Examples:

Word clouds are a classic. They are easy and fast. They work well aesthetically to grab attention, so they are good for infographics. They can be okay when accompanied by other information, visualizations, or analysis. HOWEVER, they have a number of issues:

If you do not do the additional work to clean the data

1. They do not capture synonyms, words that mean the same thing (i.e. fast, quick, speedy)
2. They do not lemmatize words, group variations of the same words (i.e. eating, ate, eat, eats)\
3. They may not remove stop words, depending on the software you use (i.e. and, on, or, the, etc.)

Regardless of cleaning the data

1. They do not capture themes
2. Related to #1, they do not show word pairs or groups which can provide important context (i.e. not cheap, costs too much, make more affordable)
3. They lack context - who said what, when, how, where, what? What was the connotation? How many times did they same person say the same thing?
4. They are prone to interpretation bias. Due to the random placement, colour, and various shapes or words, it is difficult to tell the actual difference between the size, therefore frequency, of words.