Data Communication

class:inverse

<br><br><br>
## Data Communication
### DSC365: Introduction to Data Science
#### September 15, 2026

---
## Announcements

- Lab 3: Due tonight -- if you would like extra time (so you can finish MP 1), let me know
  
- **Mini-Project 1**
  + Please update the data sets page!
  + Due **Thursday September 17, 2026 at 1:59 pm (i.e. before class starts)**
  + Presentations during class that day

---
class:inverse

.center[
## Graphical Perception
]

---
### Graphical Perception

To visually display data, information is encoded into a graph. The viewer then visually decodes this information, known as graphical perception, to gain knowledge.

In previous weeks we have discussed how to choose the right plot to visualize the data. This week we will focus on the other design choices to aid in graphical perception

---
### Keep it Simple

A general guideline is to use simplicity in the design.

---
### Show the Data Clearly

**Goal**: display the relationship between the average lifespan and lifespan standard deviation of different countries by labeling each point by sex and education level. Countries are represented by each point, with the country names also appearing

.pull-left[

<img id='imguncleargraph' src='images/unclear-graph.png' alt='Original plot that shows each observation on the same scatterplot, where different symbols are used to represent the differeng groups' =''>

<div id='modaluncleargraph' class='modal'>
  
  <img class='modal-content' id='imgmodaluncleargraph'>
  
  <div id='captionuncleargraph' class='modal-caption'></div>
</div>
].pull-right[

<img id='imgbettergraph' src='images/better-graph.png' alt='Improved visualization that facets each scatterplot by country so can reduce the clutter problem from the original plot.' =''>

<div id='modalbettergraph' class='modal'>
  
  <img class='modal-content' id='imgmodalbettergraph'>
  
  <div id='captionbettergraph' class='modal-caption'></div>
</div>
]

[(Source)](https://www.jstor.org/stable/24737226?seq=1)

---
### Show the Data Clearly: Preattentive Features

Pre-attentive features "jump out" in under 500ms

- Color, shape, angle
  - spatial location
  - movement

There is a hierarchy of features: color is stronger than shape

Articles: [Aricle 1](https://www.tandfonline.com/doi/full/10.1080/10618600.2016.1209116), [Article 2](https://www.sciencedirect.com/science/article/abs/pii/S0734189X85800049)

---
class:primary
### Find the Target

???
Find the target?

how quickly did you spot the circle?

---
class:primary
### Find the Target

---
class:primary
### Preattentive Features

.pull-left[
**Shape/Color**: Dual-Encoded

<img src="data-viz-and-communication_files/figure-html/unnamed-chunk-7-1.png" alt="Demonstrates an example of a dual encoded plot, where each different group has a different shape and color." style="display: block; margin: auto;" />
].pull-right[

**Shape/Color**: Separate Variables

<img src="data-viz-and-communication_files/figure-html/unnamed-chunk-8-1.png" alt="Demostrates an example with four groups of red circles, red trianges, green circles, and green triangles. Shows how it is difficult to quickly pick out the different groups." style="display: block; margin: auto;" />
]

---
### Show Data Clearly: Pie Charts are Bad

.pull-left[
<img src="data-viz-and-communication_files/figure-html/diamonds1-1.png" alt="Pie Chart of the diamonds dataset showing the frequency of each cut in the dataset." style="display: block; margin: auto;" />
].pull-right[

<img src="data-viz-and-communication_files/figure-html/diamonds2-1.png" alt="Bar Chart of the diamonds dataset showing the frequency of each cut in the dataset." style="display: block; margin: auto;" />
]

Anything that can be put into a pie chart can also be put into a bar chart

- Research has shown that people have more difficulty with angle measurements than length measurements [(paper)](https://www.jstor.org/stable/2288400?seq=1).

---
### Show Data Clearly: Color

+ Hue: the actual color (red, green, blue, etc.)

<br>

+ Intensity: brightness

<br>

When choosing color schemes:
  + Want mappings from data to color that are not just numerically but also perceptually uniform
  + Distinguish between sequential scales and categorical scales
  + Be conscious of what color means

[(source)](https://ieeexplore.ieee.org/document/773961)
---
### Color

.center[Our eyes are optimized for perceiving the yellow/green region]

---
### Gradients

Qualitative schemes: no more than 7 colors

Quantitative schemes: use color gradient with only one hue for positive values

Quantitative schemes: use color gradient with two hues for positive and negative values. Gradient should go through a light, neutral color (white)

[(source)](https://www.zeileis.org/papers/Zeileis+Hornik+Murrell-2009.pdf)

---
### Color Blindness

Not everyone perceives color in the same way. Some individuals have [colorblindness or color deficiencies](https://en.wikipedia.org/wiki/Color_blindness).

You can take a test designed to screen for colorblindness [here](https://eyeque.com/color-blind-test/).

Suggestions:
+ Design for a black-and-white photocopier
+ Colors
  + A monochromatic color gradient scheme where possible.
  + Suggested 2-color gradient: blue/purple - white - orange
  + Avoid: green-yellow-red signaling
  + The `viridis package` (from last week's lab) contains a set of color scales designed to span as wide a palette as possible.
  + [This website](https://colorbrewer2.org/#type=sequential&scheme=BuGn&n=3) can help pick palettes as well
+ Utilize double encoding

---
### Example: Hair Color

The data `HairEyeColor` provides information of hair eye colors in a statistics class. Suppose now we want to see the distribution of hair color in this class.

.pull-left[

``` r
hairData %>% 
  ggplot(aes(x=Hair, y=Freq)) + 
* geom_col(aes(color = Hair))
```

].pull-right[

``` r
hairData %>% 
  ggplot(aes(x=Hair, y=Freq)) +  
* geom_col(aes(fill = Hair))
```

]

---
### Example: Hair Color

It is a little weird that the color in the visualization does not match the color. Let’s try to define the color by ourselves.

``` r
ggplot(hairData, aes(x = Hair, y = Freq)) + 
  geom_col(aes(fill = Hair)) +
* scale_fill_manual(breaks = c("Black", "Brown", "Red", "Blond"),
*                   values=c("black", "brown", "red", "yellow"))
```

---
### Example: Hair Color

You may have realized that in the previous figure, the color red and brown are pretty close. For an extreme case, what if you have a reader who is color-blind? There are a lot of [research](https://jfly.uni-koeln.de/color/) on which palette then to use.

``` r
cbPalette <- c("#000000","#E69F00","#56B4E9","#009E73","#F0E442")

ggplot(hairData, aes(x = Hair, y = Freq)) + 
  geom_col(aes(fill = Hair)) +
  scale_fill_manual(values=cbPalette)
```

---
### Try it for yourself!!

Now try by yourself (or with a group) with another type of figure.

(a). Use the `iris` data below. Plot the figure to show the relation between Sepal.Length and Petal.Length.

``` r
data(iris)
```

(b). Use colors to distinguish the species. Which function you have used? fill or color?

(c). Use red to represents `virginica`, blue to represents `versicolor`, and yellow to represents `setosa`.

(d). Use the color blind friendly color to distinguish three different species

(e). Besides of the color, use different shape of the points to represents the difference (Hint: Use Google or the Help Documentation).

---
class:inverse

.center[
# Data Misrepresentation
]

---
### Common Ways to Misreprsent Data

<img id='imgcompgraphs' src='images/comp-graphs.png' alt='Illustration of the impact of misleading scales on interpretation of a bar chart.' =''>

<div id='modalcompgraphs' class='modal'>
  
  <img class='modal-content' id='imgmodalcompgraphs'>
  
  <div id='captioncompgraphs' class='modal-caption'></div>
</div>

---
### What's Wrong with These Plots?

.pull-left[

<img id='imgbadpie' src='images/bad-pie.PNG' alt='Pie chart of how people voted for president in 2016.' width='90%'>

<img id='imgbadcolor' src='images/bad-color.png' alt='Map of suicide rates by country with a bad color scheme.' width='85%'>

<div id='modalbadpie' class='modal'>
  
  <img class='modal-content' id='imgmodalbadpie'>
  
  <div id='captionbadpie' class='modal-caption'></div>
</div>

<div id='modalbadcolor' class='modal'>
  
  <img class='modal-content' id='imgmodalbadcolor'>
  
  <div id='captionbadcolor' class='modal-caption'></div>
</div>
].pull-right[

<img id='imgbadline' src='images/bad-line.PNG' alt='Line chart of Asian Hate Crime reports over time.' width='100%'>

<img id='imgbadbar' src='images/bad-bar.png' alt='Bar Chart of Number of Covid-19 cases by country over time.' width='100%'>

<div id='modalbadline' class='modal'>
  
  <img class='modal-content' id='imgmodalbadline'>
  
  <div id='captionbadline' class='modal-caption'></div>
</div>

<div id='modalbadbar' class='modal'>
  
  <img class='modal-content' id='imgmodalbadbar'>
  
  <div id='captionbadbar' class='modal-caption'></div>
</div>

]

---
### Election Maps

.pull-left[

<img id='imgelectionstate' src='images/election-state.png' alt='Election Map that colors in each state by which presidential nominees won that states electoral votes.' =''>

<img id='imgelectionpop' src='images/election-pop.png' alt='Election Map that incorporates information about population.' =''>

<div id='modalelectionstate' class='modal'>
  
  <img class='modal-content' id='imgmodalelectionstate'>
  
  <div id='captionelectionstate' class='modal-caption'></div>
</div>

<div id='modalelectionpop' class='modal'>
  
  <img class='modal-content' id='imgmodalelectionpop'>
  
  <div id='captionelectionpop' class='modal-caption'></div>
</div>
].pull-right[

<img id='imgelectoral' src='images/electoral.png' alt='Election Map that displays the number of electoral votes each presidential candidate won as hexagons, where the hexagons are arranged in a similar shape as the United States.' =''>

<div id='modalelectoral' class='modal'>
  
  <img class='modal-content' id='imgmodalelectoral'>
  
  <div id='captionelectoral' class='modal-caption'></div>
</div>

[Other Examples](https://www.flerlagetwins.com/2020/10/presidential-election.html)

]

.center[
[Source 1](https://storymaps.arcgis.com/stories/427c500cfa5f459590671a8d060c4ab4) and [Source 2](https://medium.com/swlh/the-problem-with-election-maps-443b3a941dc7)
]
---
### Visualzing Uncertainty

- Uncertainty is fundamental to data analysis and models
- Effectively communicating uncertainty in visualization is hard and is still being researched. 
- Don't present the uncertainty information without describing what it means.

.pull-left[

<img id='imgmeaninterval' src='images/mean-interval.png' alt='Visualization of a mean rating for different countries where confidence intervals are included to show uncertainty information.' =''>

<div id='modalmeaninterval' class='modal'>
  
  <img class='modal-content' id='imgmodalmeaninterval'>
  
  <div id='captionmeaninterval' class='modal-caption'></div>
</div>
].pull-right[

<img id='imghurricane' src='images/hurricane.png' alt='Visualization of different ways uncertainty information about hurricane paths can be incorporated onto a plot.' =''>

<div id='modalhurricane' class='modal'>
  
  <img class='modal-content' id='imgmodalhurricane'>
  
  <div id='captionhurricane' class='modal-caption'></div>
</div>
]

---
class:inverse

.center[
# Writing about Visualizations
]

---
### Writing About Your Figure

Your visualizations should include:
 - Clearly labeled axes
 - clear and descriptive captions
 - readable and easily understandable legends (if applicable)
 
This is *in addition to* writing about the figure in the body of the report (more on this in a moment)

---
### Labels

(a). Use plain English `\(\\\)`
(b). If there is mathematically symbols in your label, use `expression`:

``` r
x = -5:10
y = x^2
xy = data.frame(x,y)
ggplot(xy, aes(x = x, y=y)) +
  geom_point() +
* xlab("X") + ylab(expression(X^2))
```

---
### Captions

There are two ways of writing a caption:

1). In-Figure Style (for screen readers, some browsers show as mouse-over text)

Code chunk includes `fig.alt="Example of visualization created with ggplot where we added a title."`

``` r
ggplot(iris, aes(x=Sepal.Length, y=Petal.Length, group=Species)) + 
  geom_point(aes(color = Species, shape = Species)) + 
  xlab("Sepal Length") + 
  ylab("Patal Length") +
* ggtitle("Relation between Sepal Length and Petal Length")
```

---
### Captions

2). Formal report style: use descriptive caption

+ It is not included in the figure. Usually placed below the figure.
+ Starts with the word “figure” or “table” followed by a number and a colon
+ One sentence to clearly explained What and How is being compared
+ One sentence to summarize the main conclusion.
+ Any other necessary context

**Note**: label not required, but can be useful for [cross-referencing](https://quarto.org/docs/authoring/cross-references.html)
---
### Captions

What this then looks like:

---
### Captions

The caption only shows in the rendered version of documents.

Why do we need a descriptive caption? 
- The figure can stand alone, allowing the reader

How do we know we have written a good caption?
- The reader can capture the main argument without actually seeing the figure

This might seem counterintuitive, but all parts of your report, presentation, etc. work together to convey your message.
 
---
### Writing About Figures in Reports

Once you generate one figures, here is what you should do:

- Re-introduce the visualization and main argument
- Cite any statistical evidence or figure characteristics to support your argument
- State secondary argument and other information
- Contrast with the other figures if necessary
- Provide potential reasons and explain the conclusion

---
### Example: Writing About Figures in Reports

Figure 2 indicates that there is no linear relationship between iris sepal width and petal width since the scatter plot did not show a clear linear shape on the distribution of the point.

Meanwhile, the three species does have clear different distributions of the sepal width and petal width and within a species, the the sepal width and petal width may have a positive linear relationship except Setosa, which does not have a clear trend.

The relative size of different iris species is similar to what we have found in Figure 1, where Virginica tends to have the largest size among all three while Setosa is the smallest. One of the potential reasons for the different sizes may be the habitat condition while Virginica tends to grow in the places with enough sunshine but Setosa prefers relatively dark and wet environment.