Tableau study
The logical errors of your data story--overgeneralization and sample bias
Advice:
Charts to choose to show data analysis project
1. Stick to bar and line graphs. Bar charts should be used for comparing measures in different groups or categories. Things like salaries of different jobs of categories. Bar charts are appropriate for most situations where you want to show aggregated rather than raw data, and I would argue that in business presentations with general non-technical audiences, you'll most always want to be showing aggregated data. Line charts are best for looking at how values and categories vary over time, like when we were investigating whether subcategories of data-related salaries change over time. You can also use line charts to show values along another type of variable but make sure the variable has an ordered structure to it that is sequential and that has the same amount of space between them. One danger with line charts is that our eyes naturally follow the lines in the chart, and interpret them as if the points in the line have some kind of direct, ordered relationship. So when the points don't have an obvious sequentially relationship, it takes longer for us to understand what the chart is showing, and there's actually a good chance we will misinterpret the point the graph is trying to make. So remember, line charts should be made into bar charts when the ticks on the x-axis don't a specific, inherent, sequential order. 3. Pie charts can be useful when one, you are trying to communicate categories that add up to 100%. And two, you are going to highlight no more than four categories and preferably no more than two categories. 4. The fourth type of chart you can use are scatter blocks which show the relationship between two variables particularly continuous variables. Scatter plots are usually used to show raw data, although it's possible to use them to show aggregated data as well. But, I find scatter plots to be way too confusing and overwhelming for non-technical audiences. Use them when you're discussing things with your technical team or people. 5. Another type of chart you should leave out of most of your business presentations are pretty much any kind of 3D chart.
How to format slides to convey your information?--Dont't make your audiences do visual math
1. The first concept to focus on is maximizing the data-ink ratio. The term data-ink ratio was coined by Edward Tuft, a statistician and political scientist who became a well-known pioneer in data visualization. Maximize the data ink ratio basically means take out everything on the chart and everything on the slide that doesn't have a clear and unique purpose. Data ink refers to the ink that represents the actual data in a graphic. 2. Your goal should be for your audience to understand the point of your slide with as little eye movement and as little reading as possible. One of the most important issues in that regard, is always, always, always, label non-obvious axis in your graphs. This is really an important point. If you forget to label your axis, nobody will have any idea what your graph is about. Use full words unless abbreviations are extremely common. Another way to reduce the amount of work the audience has to do is whenever possible, label directly on the chart instead of using a legend. It will require less eye movement and work to understand if you don't have to go back and forth between the data and the legend, or the data and the bottom of the graph. Also note that horizontal labels are easier to read than vertical labels. So whenever possible, use horizontal orientations for your axis as well as your data labels. This might mean you have to flip the orientation of your bar graph so that the bars and labels lay flat. 3. One final set of notes about making slides readable has to do with what font you should use for the text that does have to stay on the screen. The most famous one is Times New Roman. Two others are Bodoni and Garamond. Many people suggest that the minimum size you should use is 30 points. Whenever you finish a presentation, put your slides in sorter mode and reduce their size to 66%. If you can't read them well at that size, your font is probably too small for your audience to read from the back of the room. By the way, if used 30.5 means your text will no longer fit on one page, it was probably too much text to have on one slide in the first place. Following this advice often means you end up getting rid of bullet points, instead each bullet point. If it's true that you absolutely need it, will end up in it's own slide with it's own visualization which what would be my advice to do anyway.
The logical errors of your data story--lack of controls
Always include carefully designed comparison groups that should not have the effect you were looking for in your analyses to make sure the effects you observe are due to the events you think they are due to. If a key part of your data story does not have a control group, go run some control analyses, it's the only way to make sure you are interpreting your data correctly.
The logical errors of your data story--correlation doesn't mean causation
Correlation doesn't mean causation.
Visual contrast directs where to look
I opened this week with the idea that the best way to control your audiences' decision making process is to control their attention. I also told you that one of the best way to influence people's attention is to influence what they look at. Color contrast can be a very effective way of communicating what you should pay attention to. But I did not say that color is a good way for communicating detailed information. As I told you in an earlier video, our eyes cannot detect small differences in color. However our eyes can detect large differences in color. This means color is useful for coding for categories, such as this in important and this isn't important. But color is still a very poor way to convey small differences in quantitative information, especially if multiple colors are on the screen.
storytelling- story elements
One effective way to draw your audience in is to describe the experience of a character related to your problem or location related to your problem. Using details that make your audience actually picture the person or place in their mind. Stories can feature individual characters or groups of characters. Sometimes you can even create a story where the main character isn't a person, but rather a thing like a manufacturing plant or a corrupted data set. Another particular powerful method is to tell a story about your own personal experience.
Addressing and partitioning
One way to think of addressing and partition is "direction" and "scope" respectively. Think of some aggregate calculation, like, say Average, that you might want to display in a viz. Well, do you want that average computed across the entire view in one lump? Average every year? Every month? Those would be examples of partition. The partition tells Tableau when to start and stop the calculation (scope.) Addressing is about "direction." If for example you have used the "Compute Using" dialog (say, to change from "Table Across" to "Table Down,") you were already specifying addressing. This about which elements to include in a computation, and in what order. It's about "compute on the basis of what?" So, you could think of addressing being "what to compute" and partition being "where to start and stop computing."
Calculation types in tableau
Row-level calculation, blending and aggregation-level calculations, table calculations and parameters Row-level calculations are calculations that make a different number for every single row of your data. For example, logical function like 'if' or 'case'. Blending and aggregation-level calculations mean combining two sheets together based on a common field. The larger data sheet needs to be aggregated first in order to combine with smaller data sheet. To blend, the smaller dataset needs to be primary data set, and larger dataset is secondary. And the variable name that links the two sheets need to be exactly the same to use 'automatic' combining, or we need to 'custom' Tableau to make it happen. For example, 'state' and 'work state' are not the same, so we need to customize it.
storytelling-ordering presentation slides
Storyboarding is the process of identifying the key scenes in your processor story and putting them in a logical order that conveys your message compellingly. Storyboarding is, basically, a way of having a plan for your presentation that you can articulate and communicate with others. So that you can get feedback and work in the presentation together. 1. The storyboarding process starts with you brainstorming and writing down every single insight you discover during your analysis, that you think was important at helping you arrive at your conclusion and business recommendation. Each insight needs to go on its own Post-it, index card or on its own box in your software program. Each point or detail on a Post-it should be considered a story point, that will eventually get its own graph and its own slide. Each story point should be able to be summarized in one sentence. If it takes more than that, it's too complicated and should be broken down into separate story points. Once you've gotten everything written down, you need to ruthlessly whittle down your story points until you have only the ones that are absolutely critical for justifying your recommendation for a business process change. As a rule of thumb, you should try to have no more than three main story points. And each of those main story points should have no more than three sub-points. Three items is about the limit of complexity most people can handle in one sitting. 2. Once you have your story points whittled down, the next step will be organize them according to the order which you are planning on telling them. The order should reflect what you believe is the most compelling logical argument you can give to support your recommendation. Not the order in which you actually did the analysis. If your recommendation is not going to be very controversial, I suggest starting with your strongest story point first, so that you can get your audiences buy in as quickly as possible. If your business recommendation is going to be controversial, I suggest you use a different strategy. In this type of scenario, start with your least controversial point first. The psychology literature tells us that people are more likely to be persuaded by an argument if you get them into a general feeling of agreement first.
What does ATTR() mean in Tableau?
The way ATTR() was described to me by Joe Mako is that it runs something like "If MIN([field]) = MAX([field]) Then return [field]), so it ends up being one number. Two ways it's useful: - Tableau uses everything in the Level of Detail shelf as part of the query, so doing as much aggregation as possible on those values can speed those up. If you know there will only be one value, then applying ATTR() to a dimension or measure there can make your view a bit faster. - Sometimes I have a single value for all rows in the database. Applying ATTR() to that value makes sure that Tableau won't try to do any sums or other aggregations on it.
