Checklist
I am not sure what the questions for 4.2 are. I think these are good to review any visualization you have. The story should be set before you work. What are you trying to tell? Next you should try and make this as clear as possible to make your visualization as effective as possible. To further this, you should make it as easy as possible for the readers be reducing the effort it takes for them to come to your conclusion. You should also label to make "their" findings concrete.
This grouping leads readers to conclude that living in the North is safer than living in the South, at least as far as homicide rates are concerned.
Grouping is the most powerful because it can change the story the readers walk away with. Without grouping, the readers might have the takeaway "wow the U.S. is pretty dangerous." Which is partially true. With grouping, the story becomes "wow the South is much more dangerous than the North and Western part of the country." This grouping can change the way the reader consumes the data. This can make for very large changes in the impact of the graph.
Note how Figure 4.9 draws the author’s intended story into focus by using a different color for Slovak men than for the other three groups.
Similar to how grouping can help readers pick up on the trends you want them to, bolding or highlighting data can focus readers attention where you want them to. In this graph, the line for the Slovak men data is in orange which draws you to focus on that line rather than the other three. This can be effective if you want to give some context (for credibility) but you want the reader to walk away with the concept you want them to.
Table 4.4. Visualizations with Excessive and Minimal Nondata Ink.
The simplest way to remove visual clutter from graphs is to remove background color and remove grid lines. To me, these are the most common cases of visual clutter that are easily fixed. 3D and shadow effect are also pretty bad to look at. I think they make data much more difficult to read compared to background color and grid lines. They should not be used.
This grouping leads readers to conclude that living in the North is safer than living in the South, at least as far as homicide rates are concerned.
I believe grouping is the most important way to reinforce a data story. Data can be interpreted many different ways by different people. By grouping data, you push a reader to notice the trends you want them to before they can create their own. As data visualization is convincing your reader of a certain viewpoint, this is the most powerful way.
PART II
ChatGPT seemed to make a similar graph to the one I used. Overall it was pretty good! I would make some minor tweaks but that is just on aesthetics I like (tick marks). This was faster that it took me.
Data set 4:
I would use a horizontal bar chart for this. However, I would stack the colors on top of each other to make it more readable. An alpha value of 0.5 would make this easier to read. Title: Revolution vs Standard Hemet for Concussion Downtime x = percentages y = interval color = helmet labs( x = Percentage of Players Recovered y = Time to Recover color = Helmet Type )
Data set 3:
I would use another line graph for this. The number of brands analyzed isn't that important and can be put in a caption. Title: Average Calories and Average of Sodium in Common Hotdog Meats x = calories y = sodium color = type label( x = "Average Amount of Calories" y = "Average Amount of Sodium (mg)" color = "Hotdog meat type" ) Caption: Based on analysis of 20 beef hotdog brands. 17 meat hotdogs brands, and 17 poultry hotdogs brands.
Data set 2:
I would use a line graph as it shows a trend over time. Title: Energy used in a building in Detroit and Austin x = time y = energy color = location label( x = "Month" y = "Energy used (averaged kilowatts / hour) ) A legend should be created with color.
Data set 1:
I would use a bar chart. Title: Figure 1: Percentage of Students Satisfaction with their Majors x = major y = percent label( x = "Major" y = "Percent of students "satisfied" or "very satisfied" )
Because the prompts are natural language, users can write simple, intuitive commands such as “make the bars wider,” “get rid of grid lines,” or “re-sort data from earliest to latest,” or similar.
I think this is the main issue with using GenAI to make graphs. When you use them you often don't use your own brain to make decisions and it will make them for you. These often aren't what you would have thought. I have this problem sometimes and it ends up taking more time that just doing it myself.
Table 2: Popularity of the top-selling motorcycle brands among registered motorcycle owners in Pittsburgh. Harley-Davidson is the most popular.
This is a bad caption. The first sentence is fine on its own. The second adds too much information.
Figure 4: Harley-Davidson is the most popular motorcycle.
I am not the biggest fan of this caption. It would probably need to have the data of other motorcycle brands and thus would not be descriptive of all of them. It could work if you somehow highlighted Harley-Davidson over the others but it would not be the best practice to.
Table 1: Memory usage of five different web browsers with one tab open.
This would also be a good title of a graph as it is descriptive of the two axis.
Figure 3: Levelized costs (in dollars per megawatt hours of electricity) of five different power plants.
I think this would be a good title for a graph. It seems detailed and introduces the y and x axis
Figure 2 shows the average test scores of male and female students.
Average test scores should once again be an x or y axis label instead of a caption.
Figure 1: Average GPA (on 4-point scale) by hours studied per week of students at University X. Figure 2 shows the average test scores of male and female students. Figure 3: Levelized costs (in dollars per megawatt hours of electricity) of five different power plants. Table 1: Memory usage of five different web browsers with one tab open. Figure 4: Harley-Davidson is the most popular motorcycle. Table 2: Popularity of the top-selling motorcycle brands among registered motorcycle owners in Pittsburgh. Harley-Davidson is the most popular.
This caption is too long and encapsulates two main points. Average GPA should be on the x or y axis. The hours studied per week at Uni X should be a title.
Pie graphs may be used only when percentages are reported and all values add up to 100 percent.
Pie graphs would be useful when you want to show rankings in a fast way (with a low amount of values). They become unusable when you have too many small categories. This is because they end up being so small it becomes hard to read.
Callout: Chart, graph, figure? What’s the difference?
I would use a graph when I am using important data such as crime rate or failure tolerances. This is because graphs contain exact values. Charts on the other hand are simplified. This may be useful in situations where the exact number is not as important as the trend. This could be GDP per capita of 5 countries. The difference between 55k and 54.8k is not as important as the trends between the five countries.
Leaving out important context or information that would change readers’ understanding of the data
I plan on using this when working on my team's milestone. I think this can be one of those things that very tempting in academia. Because you want your research to have some conclusion, you feel like you should leave out some context to strengthen your argument.
4. Carefully word claims to avoid exaggeration
I think this is the most common way I have seen authors lose credibility. It's really common to use correlation to imply causation online. I see it very often in pop culture/science online accounts.
Leading readers to believe that small differences are actually large (or vice versa)
Small increases in exponential events can lead to large gains later on. Small gains in investment accounts can lead to massive gains later on in life. Large chunks of battery percentage on my laptop don't really make a different. I generally only use 30-40% of my laptop battery a day. If I lose like 20-30% I wont notice as my laptop won't die.
Our data show that the new interface is less popular with users than the old one was.
Our data shows that the new interface is less popular than the previous version. We should either roll the update back or investigate what made the previous version more appealing and add those changes in our next update.
Without words guiding them to what is most important, readers can misread your data or fail to grasp the data’s significance. And if you provide words without visuals, readers will question your credibility and ask for more detail.
I think both are needed because they make up for what the other lacks. Visuals can contain a lot of information but often it can be difficult to get the argument across on its own. Text can guide your reader to your point but aren't as powerful as visuals when it comes to convincing.
Are any of your revisions more or less trustworthy than others? Do any cast doubt on your credibility?
I think the woman's one is lowers my credibility as I did not give context (just saying it was about double). I think the other two are both as trustworthy as each other.
How might you rewrite it to minimize the importance of depression?
"About 85% of the population will not experience depression in their lifetime"
What if you wanted to encourage more research into men’s experiences of depression?
"A little more than one in ten men will experience depression in their life"
How might you rewrite this statistic to encourage a woman’s organization to support depression research?
"Women are nearly 2x as likely as men to experience depression in their lifetime"
The exact words that a dissatisfied user chooses to characterize an experience with a website can often be more meaningful than numerical data on the number of failed submissions or the number of views.
Qualitative data is often more consumable that quantitative data. This means it can be easier/better when presenting. Qualitative data can be expressed quantitatively (ranking something).
In both cases, we have Steve communicating that the test was difficult, but the two versions produce very different impressions of Steve — just as the two different ways to describe the likelihood of winning the raffle produce different impressions of the odds.
Framing your argument in a dishonest way can cause your credibility to take a hit. Audience can become more engaged if you frame it well. Your purpose can influence how you frame data and statistics as you need to know your what you are trying to say before you get the data and/or utilize statistical methods. As written earlier audience and credibility are highly linked. Both need each other to be effective. They can conflict if one is done poorly.
These three considerations of purpose, audience, and credibility are intimately connected, and decisions we make about one of these considerations will affect the other two.
The three are pretty intertwined. When presenting your audience, you need to establish credibility so they listen to you. To establish your credibility, you need to know your audience! Its hard to do either of these things without having a reason (purpose) to do it. Prioritizing one over another can have its time and place. For example if you are presenting data that is controversial to a group, you need to work hard on credibility. If you are trying to cause policy to change you need to focus on your purpose.
In other words, the same number presented two different ways can have very different effects on the audience.
The purpose of the data portrayal influences how it is presented. If the author wants the audience to think a certain way, it can be done with data that isn't very strong.