How to Graph Badly
In “How to Graph Badly or What NOT to Do”, the author breaks down different examples and explanations of how to not represent data in graphical visualizations. According to the author, much of the skill in creating good, helpful viusalizations lies in not making mistakes. There is something called “Chartjunk”, which is a term coined by Edward Tufte for adding useless features with no informational content of graphical visualizations. If a simple graph is able to tell the story of that data, do not fill it with extraneous decoration, because the simplicity and readibility of the graph will be taken away. The focus from the data is then being used to figure out what is important to look at. Fonts are also very important to making a graph, because there is danger in illegibility. Make sure when making a graph with labels, the reader is able to read the ledger and manuscript without trouble. In addition, many chose to shade their graphs. The cross-hatching is eye catching, which can be distracting and pull away from the heart and story of your data. 3-D graphics can be useful and helpful, with the right data and setting. When using 3-D graphics, there must be extra care given to make sure the graphs are able to be percieved correctly. In addition, the color of the graphs are important because not only does color on graphs increase price in storage, but it also makes graphs very clear of what it is trying to relay. When looking at your data, there is no need to draw conclusions and correlations from data sets if there is no clear causal relationship. When creating visualizations, varying the graph types, highlighting key features and combining closely related graphs will help your graph look and read better when there is a lot of complicated information involved. There is also a list included for rules for bad graphs, specifically 12 rules. Data density should be minimized to justify concepts of information as well as illustrate a concept well. It needs to be understood that when words can be as clear as a graph, chose to use words over graphing something simple or complex. Also, the author says to split graphs into multiple panels in order to reduce reading error. There is something called “visual metaphor” where it stands for the relationship between graphical elements and the data. Incosistent use of a visual metaphor is more common than not. Also, in a graph it is difficult to explain and describe the background of the graph because it is a graph. The most important part of a good graph is that is shows enough curves. It is often a bad idea to over complicate simple graphs. Keep graphs minimal and do not overcomplicate using photos and designs. In all graphs, make sure that the labels included, if they are included, are legible, complete and presice. Also, make sure the placement of the labels are suffient in the way that it does not block any of the data points or overcomplicate the graph. With colors, it is often helpful to use different shadings to differentiate different graphical elements one from another. This can also be a bad thing, and make readers focus on one bar or data set more than the others. Overall, you must know what is comprehensible and familiar for the intended audience.
The Gospel According to Tufte
This article talks of graphical simplicity and what data-ink is. The concept of data ink is more than just “simplicity”, it can be elaborated into 5 maxiums, according to Tufte: show the data, maximize the data-ink ratio, erase non-data-ink, and revise and edit. One of the most important parts of the rules is to always show the data. Go into making graphs with a preconcieved notion to have a topic sentence or general statement. Knowing the concept going into creating the data allows for a firmer understanding of the topic at hand. Secondly, maximizing the data-ink ratio should be utilized because you don’t want to be underemphasize data, so you should make graphs with easy to read labels in order to make a lasting impression. The author warns us to be skeptical of grid lines, as they may confuse the audience. There are only a few reasons to use grid lines in a graph, when the figure is a nomogram or when the author expects the reader to carefully study a curve and pick out heights, which would be helpful in supplying visual cues. Also, if the grid lines carries a message, there should be use of the gridlines then. The author talks about a different simplification - half framing. Half framing allows more whitespace and emphasizes the data-ink theme to strive for simplicity. In addition, to illustrate bar graphs, one may chose to redesign and go for a more simplictic approach, but users and viewers may be unfamiliar with the new format. Sometimes, you may want to graph both halves of a symmetric object to convey the parity of the object. Humans prefer symmetry, and this is one way to provide for that. Keep in mind that not everything has to be shown with a graph, it can be shown with a table as well. Eliminating the cluttered things on the graph may take away from the story of the graph, so turning the graph into something different would perhaps allow for more understanding. As stated earlier, it may be good to go into creating a graph with a “topic sentence” in mind. In addition to this, do not be afraid to revise and edit your scientific visualization - it is important to do such a thing. One also has to be cognisant to not overwhelm the reader with too much data in one graphic visualization. Design high density graphs with the reader in mind. and design the graphs so they present the themes and goals of the study. In addition to high density graphs, be aware of the shrink principle - graphs can be shrunken down and turned into a single composite figure. There are ways to convey data, and for example, a drawing that is a graph which also functions as a table may be able to convey a lot of important information to the reader. Multifunctioning graphs may be cool, but not always as effective as needed. There may be things that are redundant, unneccessary, or extraneous to the topic at hand. You can also include a “small multiple” which is an animation on a page. They have to differ by a small amount frame by frame. This is a collection of mini illustrations, shown as just one. This may not be used on a diversity of multipanel figures because the diversity is not goof for this specific visualization. 1+1 = 3. This is an important phrase to data visualization because you have to understand that elements of a grpah interact with one another. Within something called “layering”, there is the seperation of different layers based on space, color, style, and tyoe font. Color and greyscale are important tools for seperation, because it actually seperates by color, which would be to layer the colors over one another - although this may be expensive, it is valuable. “Seperation” could also be physical proximity, and is especially important for labeling on a graph. This leads into labeling - all labeling should be clear and explicit - it should not be too distracting. “Using space as a substitute for time is not the only strategy for “compactifying dimensions”. Another is to ignore irrelevant dimensions." There is also the use of inset graphs, which illustrates spacial struccture and a particular time. There is also the extrenal aspect ratio of a graph, and thagt is when the ratio of its width to its height as it appears on the printed page. It is desirable because humans prefer it visually, because it is not “dangerous”. Also, in a certain software, it is unable to wrap text around illustrations, so there is no wasted white space. Color can emphasize wrong elements of the graph, so limit yourself to specific color palletes- it might be bright and primary colors or it might be softer tones that are easier to understand. But using too many colors should not be used, because they will be indistinguishable and will not convey the right meaning. You also have to be aware that color blindness affects a small percentage of the population, so a rainbow pallete of colors will be problematic. Parallel images are easy to grasp because not only is it favorable, but it is simple and can be easily comparable. Overall, be clear, precise, modest and understand who the readers and viewers of the data graphics are in order to create successful and informative graphs.