Supplementary Materials for ASSETS’21 Experience Report
Interdependent Variables: Remotely Designing Tactile Graphics for an Accessible Workflow

Overview

These supplementary materials offer examples of tactile graphics that co-authors Lilian de Greef and Cynthia Bennett created while developing their accessible workflow around data representations. Our full set of supplementary materials contains this document and two folders: a folder titled "EmbossableGraphics" containing TactileView files of embossable graphics, and a folder titled "ImagesForHTMLFile" containing images for the file you're reading right now. In the next section of this document, we share an overview for how to emboss the graphics in "EmbossableGraphics". We subsequently share photographs of example reference sheets and tactile graphics we developed, organized by handmade then embossed materials.

Embossing the Tactile Graphics

The "EmbossableGraphics" folder contains graphics we created, both for tactile reference sheets and for data representations like the ones detailed later in this document. These files require TactileView software to use and are specifically designed for the ViewPlus Columbia 2 embosser and 11.5 x 11 inch paper. They may not be compatible with other embossers. When embossing these graphics, refer to the ViewPlus Columbia's User Manual. Remember to set the embosser to use 11.5 x 11 paper, and Tactile Graphics (as opposed to text). Also remember to adjust printer settings to use the VP Columbia printer, "Actual Size" (instead of "shrink"), and "Best" graphics quality (which is under printer properties / Tiger tab). Most of the graphics were designed for landscape orientation.

Handmade Materials

Handmade Tactile Reference Sheet

As discussed in the report, we created a reference sheet of different tactile materials. Figure 1, below, offers a larger image of this reference sheet. Figure 2 shows the backside of this reference sheet.

A portrait card stock sheet of paper titled “Texture Inventory” in both ink and braille. It features lines, dots, and other pieces of different tactile materials on an 8.5x11 sheet of card stock paper. It labels them as the following: Elmer's glue, 3D paint, Soufflĕ Pen with x1, x2, and x3 for the number of layers, Glitter Glue, Hemp Cord, Hot Glue, Masking Tape, Blue Tape, Washi Tape (shiny), Washi Tape (Cherry Red), Washi Tape (Glitter Silver), Shipping packaging tape, Duct Tape, Medical Tape, Sports Tape, Strapping Tape, and Stick-on items, including googley eyes, mini pom-poms, plastic rhinestones, and different types of surface protectors.
Figure 1: A handmade tactile reference sheet featuring 24 different materials for creating different textures. For some materials, the reference sheet includes examples of both applying one layer and two or more layers.
A cardstock sheet of paper with 8 different lines drawn with 3D puff paint. They are: a solid line, dotted line, dashed line, two parallel lines, two parallel lines drawn more closely together, a line with miniature loops, a line with perpendicular tick marks, and a line with small squiggles. The photo also shows the back side of the hemp cord that was stitched into the paper.
Figure 2: The back side of the handmade tactile reference sheet, featuring different styles of lines created with puff paint.

Handmade Tactile Graphics

In this section, we showcase a subset of our handmade tactile graphics. One is a tactile version of Anscombe’s Quartet (Figure 3), a classic example of how data tables or statistical summaries cannot replace the utility of graphical data representations [2]. While the basic statistics for the data in these four charts like the mean x value or correlation are nearly identical, graphical representations quickly reveal different underlying patterns.

A handmade tactile graphic of a scatterplot, titled “Scatter-03 - Anscombe’s Quartet I” via a braille and ink label at the top. The scatterplot covers most of a landscape 8.5x11 sheet of card stock paper with the top right corner cut off. Straight lines of puff paint mark X and Y axis lines, and faint hand-embossed lines traced with red ink mark gridlines. The X-axis braille and ink labels read from 4 to 18; Y-axis labels read from 4 to 12. The scatterplot has 11 data points made of stick-on fabric circles, commonly used to protect hard flooring from the feet of furniture scraping it. It also has line of best fit with a positive trend, made of hemp cord glued onto the paper. They generally trend with the line of best fit; 5 of the points intersect with the line and the remaining 6 are either above or below the line. They are variably spaced apart, ranging across the X axis from 4 to 14, and across the Y axis from 4.26 to 10.84. A handmade tactile graphic of a scatterplot, titled “Scatter-04 - Anscombe’s Quartet II”. It is identical to the other scatterplots in Figure 3, both in materials and content, with the exception of the title and the arrangement of its 11 data points. Like the other scatterplots in Figure 3, the X-axis labels read from 4 to 18; Y-axis labels read from 4 to 12. Unlike the other scatterplots in Figure 3, these points are evenly spaced in a way that forms a curved line. Starting from (4, 3), the line of points slope upward, first at a steep angle then gradually leveling off to the top of the curve around (11, 9). It then curves back downward and ends at (14, 18). This curve of points starts below the line of best fit, goes above the line of best fit around (6, 6), and drops below the line of best fit again around (12, 9). This line of best fit is the same as the other scatterplots’ in Figure 3. Like the other scatterplots in Figure 3, this graphic covers most of a landscape 8.5x11 sheet of card stock paper with the top right corner cut off. Straight lines of puff paint mark X and Y axis lines, and faint hand-embossed lines traced with red ink mark gridlines. The data points are made of stick-on fabric circles, commonly used to protect hard flooring from the feet of furniture scraping it. The line of best fit has a positive trend, made of hemp cord glued onto the paper.
A handmade tactile graphic of a scatterplot, titled “Scatter-05 - Anscombe’s Quartet III”. It is identical to the other scatterplots in Figure 3, both in materials and content, with the exception of the title and the arrangement of its 11 data points. Like the other scatterplots in Figure 3, the X-axis labels read from 4 to 18; Y-axis labels read from 4 to 12. Unlike the other scatterplots in Figure 3, 10 of the data points are evenly spaced in a way that forms a gently sloping straight line. This line starts above the line of best fit with a point at around (4, 5.5), gently slopes upward, crosses the line of best fit around (6, 6), and continues its gentle slope under the line of best fit until its last point around (14, 9). There is a gap between the last 2 points as if a point were missing. The 11th data point is far above this gap, at around (13 ,13). Like the other scatterplots in Figure 3, this graphic covers most of a landscape 8.5x11 sheet of card stock paper with the top right corner cut off. Straight lines of puff paint mark X and Y axis lines, and faint hand-embossed lines traced with red ink mark gridlines. The data points are made of stick-on fabric circles, commonly used to protect hard flooring from the feet of furniture scraping it. The line of best fit has the same positive trend as the other scatterplots’ in Figure 3, and is also made of hemp cord glued onto the paper. A handmade tactile graphic of a scatterplot, titled “Scatter-06 - Anscombe’s Quartet Iv.” It is identical to the other scatterplots in Figure 3, both in materials and content, with the exception of the title and the arrangement of its 11 data points. Like the other scatterplots in Figure 3, the X-axis labels read from 4 to 18; Y-axis labels read from 4 to 12. Unlike the other scatterplots in Figure 3, 10 of the data points all have the same X value of 8, and are so close together that the tactile markers for these points literally stack on top of each other between the Y values 5.25 and 9. The 11th data point is further away, at (19, 12.5), and intersects with the line of best fit. Like the other scatterplots in Figure 3, this graphic covers most of a landscape 8.5x11 sheet of card stock paper with the top right corner cut off. Straight lines of puff paint mark X and Y axis lines, and faint hand-embossed lines traced with red ink mark gridlines. The data points are made of stick-on fabric circles, commonly used to protect hard flooring from the feet of furniture scraping it. The line of best fit has the same positive trend as the other scatterplots’ in Figure 3, and is also made of hemp cord glued onto the paper.
Figure 3: Tactile representations of Anscombe’s quartet [2]. The data in each chart have identical mean and variance for x values, and nearly identical mean and variance of y values, correlation, linear regression, and coefficient of determination.

Figure 4 shows how graphical representations can highlight different properties of the same data—another example of how important it is to use tactile graphics in our workflow instead of relying on data tables.

A handmade tactile graphic of a grouped bar chart, titled “Grouped Bars-05 - Salaries 1, Figure 7 a from Burns 2010 Paper” via a braille and ink label at the top. The bar chart covers most of a landscape 8.5x11 sheet of card stock paper with the top right corner cut off, and compares female and male salaries for different industries. Faintly-tactile black gridlines demarcate a graph, with Y-axis labels for salary from 0 to 80,000 in increments of 20,000. X-axis labels are CS (meaning computer/math science), EN (meaning engineering), LS (meaning life science), PS (meaning physical science), and SS (meaning social science). Groups with 2 bars each extend up from each X-axis label; male salaries demarcated with glittery tape are higher than the female salaries demarcated with duct tape in each industry. A handmade tactile graphic of a grouped bar chart, titled “Grouped Bars-06 - Salaries 2, Figure 7b from Burns 2010 Paper”. It is identical to the previous grouped bar chart in Figure 4, both in materials and content, with the exception of the title and the order for arranging its X-axis labels and bars. Like the previous grouped bar chart, glitter tape and duct table demarcate male and female salaries, respectively. Unlike the previous chart, all the male salaries are on one side of the chart and all the female salaries are on the other side of the chart. The X-axis labels repeat for each of these two groups. Within these groups, bars for each industry are sorted by salary in order of lowest to highest. This ordering of bars makes it harder to compare male versus female salaries and easier to compare the salaries between the different industries, lending itself to showing a ranking of which industry offers the highest salaries.
A portrait sheet of card stock with the right corner cut, containing mostly braille and ink text. A title at the top reads: “legend for grouped bars-05 and 06 salaries”. The beginning of the page reads “legend: female salaries, male salaries” paired with samples of duct tape and glitter tape to match which symbolizes what. Underneath, separate lines of text pair acronyms with their phrases, via the text “CS: computer/math sci., EN: engineering, LS: life sci., PS: phys. sci., SS: social sci.”
Figure 4: A set of two tactile grouped bar charts and their legend, based on figures in [3]. Both charts represent the same underlying data, but the different ways they organize the data make different patterns stand out: one of the most salient features of the first chart is a comparison between female and male salaries, which is more difficult to perceive in the second chart. Meanwhile, the second chart makes it easier to rank and compare the relative salary levels for different industries, which is more difficult to perceive in the first chart.

In our report, we highlight how graphical representations can offer more nuance that verbal descriptions. One example is how accessing the graphic of a line chart described as “volatile” created an “ah hah” moment for understanding just how volatile the line is, both quantitatively and qualitatively. Figure 5 shows two contrasting examples of such line charts.

A handmade tactile graphic of a highly volatile line chart, titled “Line-03 - Highly Volatile, Figure 3 from Moraes 2013 paper”. Braille and inked text under title read, “ocean levels rising. Annual differences from Seattle’s 1899 sea level, in inches.” The line chart below it takes up most of the 8.5x11 sheet of card stock. The X axis labels years from 1900 to 2000, and the Y axis labels sea level difference from -1 to 10 inches. A bold, black, tactile line made of puff paint meanders over faintly hand-embossed gridlines, from just before 1900 at 1.97 inches to 2000 at 8.9 inches. Additional labels to mark these Y axis values. As this line travels along the X axis, it zig-zags erratically along the Y axis, varying from one year to the next by as little as a fraction of an inch or as much as 5 inches of sea level, the latter of which is nearly half the height of the chart. A handmade tactile graphic of a relatively smooth line chart, titled “Line-02 - Durango Sales, Figure 2 from Moraes 2013 paper”. Braille and inked text under title read, “Declining Durango sales.” The line chart below it takes up most of the 8.5x11 sheet of card stock. The X axis labels years from 1997 to 2006, and the Y axis labels sales from 0 to 200,000. A bold, black, tactile line made of puff paint stretches across faintly hand-embossed gridlines. As this line travels along the X axis, it changes smoothly and mostly predictably along the Y axis, changing directions between increasing and decreasing only 3 times. Approximately, it starts at around (1997, 10,000), increases to (1999, 190,000), decreases to (2002, 105,000), increases to (2004, 135,000), then decreases to end at (2006, 70,000).
Figure 5: Two tactile line charts based on figures in [5]: one with high volatility and one with low volatility.

Verbal descriptions alone also could not feasibly communicate all the rich details of complex charts without them becoming too lengthy or complicated. Figures 6 and 7 showcase examples of charts with more complexity than the above figures. Each member of our team, both sighted and blind, contributed different observations on patterns or specific data points when we referenced this chart during our brainstorms and explorations.

A handmade tactile graphic of a stacked bar chart, titled “Bar-03 - Screen Time. One day of phone screen time”. A braille and ink legend under title associate the different textures of tape: glitter tape for productivity, sports tape for other, soft velcro tape for reading & reference, and packing tape for “some unlabeled category”. In the chart below it, the X axis labels read from 00 to 18 for hours of the day, and the Y axis labels read from 0 to 60 minutes. Each bar on the chart is made out of segments of different materials. For example, the bar representing screen time from 7 AM to 8 AM is made of, from bottom to top, a small sliver of glitter tape, a slightly larger piece of sports tape, then a piece of soft velcro tape. The length of each segment represents the amount of screen time for its associated type of activity, whereas their combined length represents the total amount of screen time for that hour. The chart has 10 bars of varying heights, many of which have multiple segments, within the X axis interval representing 7 AM to 5 PM. Like with many of the other handmade tactile graphics, this graphic covers most of a landscape 8.5x11 sheet of card stock with the top right corner cut off.
Figure 6: An example of a chart with more complexity: a tactile stacked bar chart based on a Screen Time chart in iOS Settings, using fictitious data. It represents the four series with glitter washi tape, sports tape, soft velcro tape, and packing tape.
A handmade tactile graphic of a multi-series line chart, titled “Line-04 - Pandemic Needs Part 2”. The X axis labels read dates from 1/1 to 8/1. The Y axis, titled “relative change in needs in log scale (between pre-pandemic baseline & each day)” and ranges from -0.5 to 2. Faintly embossed horizontal gridlines mark regular intervals along the Y axis, whereas similar vertical gridlines mark irregular intervals along the X axis and are labeled with letters A through I above the chart, to be referenced in a separate key elsewhere. The chart contains 5 bold lines made of different textures or patterns. They travel complex paths that sometimes cross with each other, offering many different opportunities for finding patterns in the data. Like with many of the other handmade tactile graphics, this graphic covers most of a landscape 8.5x11 sheet of card stock with the top right corner cut off. A braille and ink legend for the previous chart in this figure, made out of a portrait sheet of card stock with the right corner cut, containing mostly braille and ink text. A title at the top reads: “Line-04 - Pandemic Needs Part 1”. Text under the title reads, “See Numbers document for key of letters”. The legend beneath it maps different textures or patterns of lines with a phrase for their particular line in the previous chart in in this figure. A line of puff paint crossed with many perpendicular tick marks maps to “Physiological”. A dotted line of puff paint maps to “Safety”. Two close parallel lines of puff paint maps to “Love & Belonging”. A line of glitter glue maps to “Cognitive”. A simple single line of puff paint maps to “Self actualization”.
Figure 7: Another chart with more complexity: a tactile multi-series line chart and its legend, based on a figure in [6].

We also report that handmade tactile graphics can express some charts more clearly than modern-day embossers. One such example is the heat map in Figure 8.

A handmade tactile graphic of a heat map, titled “Heatmap-01 - Movie Ratings” on a portrait sheet of card stock with the right corner cut. The second line of braille and ink reads, “legend: count of records”. Below it is a mapping between different numbers and different heights of thickly stacked tape. There are 10 different heights, with numbers associated with every other height. From tallest to second lowest height, the numbers read 150, 120, 90, 60, and 30. The chart below this legend has an X axis titled “IMDB Rating (binned)” with values ranging from 1 to 10, and a Y axis titled “Rotten Tomatoes Rating (binned)” with values ranging from 0 to 100. Lines of puff paint mark the axis lines and faintly hand-embossed gridlines divide the chart into 100 squares. Each square is either empty, representing a data value of 0 records, or frames a stack of thick tape, representing the value associated in the legend. The highest concentration of records follows an upward trend, increasing from around (4, 0) to (9, 100). Most of the remaining area of the heat map is filled with lower numbers of records. A side view of the same heat map, showcasing its three-dimensionality. Some sections are as tall as half an inch.
Figure 8: A tactile heat map based on [1], photographed from two different angles. It uses layers of thick medical tape to map different three-dimensional heights to the value in each of the heat map’s cells, finished with a layer of moleskin padding (which is another type of medical tape) to reduce friction for gliding fingers across the surface.

Embossed Materials

The figures in this section represent a subset of the graphics we embossed, each pictured with both a digital representation and a photograph of the embossed output. Each of the figures (and more) are also available through the EmbossableGraphics.pdf file included in the supplementary materials.

The embosser punches dots into paper based on where the digital representation has gray pixels, using the tallest dot height for black pixels and subsequently shorter dots for lighter shades of gray. The Embosser ignores blue pixels, which enables annotating the files with print text.

Embossed Tactile Reference Sheets

We created multiple reference sheets that compare different lines, fills, and dot heights (Figures 9 and 10).

A digital version of an emboss-able file, titled “Line Heights & Thicknesses” in braille and print at the top. Under the title is the text “row: height, column: thickness”. A grid of short horizontal line segments cover the rest of the page, with rows labeled numbers 1 through 8 and columns labeled as 5, 10, 15, and 20. Line segments in each row have increasingly darker shades of gray, corresponding with the options for 8 different dot heights. Line segments in each column have different thicknesses. Each row-column pair has both a solid and a dotted line. A photo of the previous reference graphic, “Line Heights & Thicknesses”, embossed onto a 11x11.5 sheet of paper. The increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
A digital version of an emboss-able file, titled “Shape Fills and Borders” in braille and print at the top. Under the title is the text “row: fill height, column: border height diff”. A grid of repeats shapes cover most of the page, with rows labeled numbers 1 through 7 and columns labeled as “none” followed by numbers 0 through 7. The shapes for each row have increasingly darker shades of gray, corresponding with the options for the first 7 different dot heights. The shapes for each column have increasingly dark borders, where the number labeling a column corresponds to how many increments darker the border is from the fill. An additional pair of shapes toward the bottom left corner of the page are completely black, corresponding to the tallest dot height, and are prefaced with the text “fill height 8, none, 0”. The very bottom of the page reads “note: “none” means no border and 0 means same height as fill”. A photo of the previous reference graphic, “Shape Fills and Borders”, embossed onto a 11x11.5 sheet of paper. The increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
Figure 9: Top: an embossed grid of solid and dotted lines with different widths and dot heights. Bottom: a grid of the same set of shapes featuring 90-degree, angled, and curved edges, but with different combinations of border and fill dot heights.
A digital version of an emboss-able file, titled “Adjacent Areas with Different Heights” in braille and print at the top. Under the title is the text “Numbers represent dot height of rectangle”. The rest of the graphic is a rows of borderless gray filled rectangles touching each other with numbers above them, and includes every possible combination of adjacent dot heights. The first row of adjacent rectangles has 8 rectangles ordered from lightest to darkest, labeled with numbers 1 through 8. The next row has two sets of 4 adjacent rectangles each, the first labeled with numbers 1, 3, 5, 7, and the second labeled with numbers 2, 4, 6, 8. The third row has three sets of adjacent rectangles, the first labeled 1, 4, 7, the second labeled 2, 5, 8, and the last labeled 3, 6. The remaining rows contain pairs of adjacent rectangles, paired off with labels 1 and 6, 2, and 7, 3 and 8, 1 and 7, 2 and 8, and 1 and 8. A photo of the previous reference graphic, “Adjacent Areas with Different Heights”, embossed onto a 11x11.5 sheet of paper. The increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
A digital version of an emboss-able file, titled “Circles and Angled Lines” in braille and print at the top. Under the title is the text “Filled circles by diameter (2 of each):”. Below it is a row of numbers 5, 10, 15, 20, 25, 30, 35, 40, and below each number, two black circles with sizes corresponding to these numbers. The first two circles, associated with number 5, are the size of single braille dots, whereas the last two circles, associated with number 40, are about 8 times as wide. Partway down the page, text reads “Hollow circles by border thickness and diameter” followed by a grid with 3 rows labeled “Border: 5, 10, 15”, and 8 columns labeled “diameter (in inches): .1, .2, .25, .35, .5, .6, .75, 1”. At the intersection of each row and column is a black hollow circle, with an increasingly thick border along the rows and an increasing diameter along the columns. The bottom third of the page starts with the text “Angles and curved lines for 2 line thicknesses”. Below it are two sets of identical black lines, the second set using thicker lines than the first. Each set contains 8 straight lines at differently sloping angles and a spiral. A photo of the previous reference graphic, “Circles and Angled Lines”, embossed onto a 11x11.5 sheet of paper.
Figure 10: Two more embossed tactile reference sheets. Top: a set of filled rectangles touching each other for tactile reference of every possible combination of adjacent dots heights. Middle: different sizes of filled and hollow circles, as well as lines at different angles. Bottom: a test for overlapping areas of one dot height with markings of another dot height.

Embossed Tactile Graphics

Figures 11–12 show a few examples of our embossed tactile graphics. Figure 12, in particular, is an example of making two variations based on the same chart, which is much faster and easier to do with an embosser than making tactile graphics by hand.

A digital version of an emboss-able line chart, titled “Fantasy Magic Candy beans Stock” in braille and print at the top. Under the title is the text “Note: company and prices are fictitious”. The X axis, titled “Trading date”, has the labels Jan, March, May, and June. The Y axis, titled “Trading price per unit stock ($)” has labels 55, 75, 95, 115, and 136. Two medium-gray lines mark the axis lines, and very pale gray lines extend from the axis labels to form gridlines. On this chart, a thin black line starts toward the bottom left corner of the chart and meanders its way toward the top right corner of the chart. A photo of the previous tactile graphic, “Fantasy Magic Candy beans Stock”, embossed onto a 11x11.5 sheet of paper. Increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
A digital version of an emboss-able bar chart, titled “Step Count” in braille and print at the top. Under the title is the text “one day’s imaginary step count from 12am to 7pm”. The X axis, titled “Hour of the Day”, has the labels 12 am, 6 am, 12 pm, and 6 pm. The Y axis, titled “Number of Steps (in thousands of steps)” has labels 0, 1, 2, 3, and 4. Two medium-gray lines mark the axis lines, and very pale gray lines extend from the axis labels to form gridlines. Thin black bars represent data on the chart, with bars of varying height between 8am and 8pm. A photo of the previous tactile graphic, “Step Count”, embossed onto a 11x11.5 sheet of paper. Increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
Figure 11: Examples of embossed tactile graphics: a line chart and a bar chart.
A digital version of an emboss-able single-series scatterplot, titled “Penguin flipper length versus body mass from penguins on the Palmer Archipelago, Antarctica” in braille and print at the top. The X axis, titled “Flipper Length (mm)”, has 6 labels ranging from 180 to 230. The Y axis, titled “Body Mass (kg)” has 8 labels ranging from 2.5 to 6. Two medium-gray lines mark the axis lines, and very pale gray lines extend from the axis labels to form gridlines. 48 filled black circles data points on the chart, loosely following a positive trend, going from the bottom left to the top right of the chart. A photo of the previous tactile graphic, the single-series version “Penguin flipper length versus body mass from penguins on the Palmer Archipelago, Antarctica”, embossed onto a 11x11.5 sheet of paper. Increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
A digital file for embossing the same scatterplot as the previous scatterplot on penguins, except as a multi-series chart. It has the same title, “Penguin flipper length versus body mass from penguins on the Palmer Archipelago, Antarctica” in braille and print at the top. Under the title is the text “Legend (species) followed by pairs of shapes and penguin species names: a filled circle for “Adelie”, a cross for “Chinstrap”, and a hollow square for “Gentoo”. The axes, gridlines, and location of data points are identical to the previous scatterplot in this figure with the same title. However, this scatterplot represents data points with the shapes in the legend instead of only circles. 21 filled black circles, representing the Adelie species, cluster primarily in the lower left quadrant of the chart. 10 crosses, representing Chinstrap, scatter around the middle and partway toward the lower left of the chart, partially overlapping with the cluster of circles. 17 hollow squares, representing Gentoo, cluster in the upper right corner of the chart. A photo of the previous tactile graphic, the multi-series version “Penguin flipper length versus body mass from penguins on the Palmer Archipelago, Antarctica”, embossed onto a 11x11.5 sheet of paper. Increasingly darker shades of gray in the digital version correspond to taller tactile dots in the embossed version.
Figure 12: Two embossed scatterplots based on a subset of the Palmer Penguins dataset [4]. Both show flipper length versus body mass for the same penguin data, except that the second chart also marks the species associated with each data point. Modifying the digital file of first chart into the second chart is much faster than re-creating them by hand.

References

  1. 2D Histogram Heatmap. Retrieved June 22, 2021 from https://vega.github.io/vega-lite/examples/rect_binned_heatmap.html
  2. Francis J. Anscombe. 1973. Graphs in Statistical Analysis. The American Statistician 27, 1 (1973). DOI: 10.1080/00031305.1973.10478966
  3. Richard Burns, Sandra Carberry, and Stephanie Elzer. 2010. Visual and Spatial Factors in a Bayesian Reasoning Framework for the Recognition of Intended Messages in Grouped Bar Charts. Visual Representations and Reasoning.
  4. Allison Marie Horst and Alison Presmanes Hill and Kristen B Gorman. 2020. palmerpenguins: Palmer Archipelago (Antarctica) penguin data. R package version 0.1.0. https://allisonhorst.github.io/palmerpenguins/. DOI: 10.5281/zenodo.3960218.
  5. Priscilla S. Moraes, Sandra Carberry, and Kathleen Mccoy. 2013. Providing access to the high-level content of line graphs from online popular media. Proceedings of the 10th International Cross-Disciplinary Conference on Web Accessibility - W4A 13 (2013). DOI: 10.1145/2461121.2461123
  6. Jina Suh, Eric Horvitz, Ryen W. White, and Tim Althoff. 2021. Population-Scale Study of Human Needs During the COVID-19 Pandemic. Proceedings of the 14th ACM International Conference on Web Search and Data Mining (2021). DOI: 10.1145/3437963.3441788