notes on Wainer (2009)

Howard Wainer 2009 Graphic discovery: a trout in the milk and other visual adventures. Princeton NJ, Princeton University Press.

The book offers a stimulating discussion of data visualization, and I especially appreciate the wide range of examples from history and the author's own research. This page consists of personal notes on the exhibits from the book. You are welcome to add to and argue with them!

Some of the questions below are rhetorical, others have answers, but it's not always clear which is which.

References are to figure numbers unless noted P for Page or T for Table.

EXHIBIT COMMENTS

1,2 V-axis should be from 0 to 20 in steps of 5 - abbreviated (0, 20, 5). The minor ticks are unnecessary, especially in a non-time axis.
As always in a time series, the V-axis annotations (title and tick labels) could be to the right.
Show either 9 minor H-axis minor ticks (one for each year) or none at all.
The exhibit could use a clear title instead of the V-axis label.
Wainer's own graphics appear to be drawn by hand, or at least to have a slightly crude look (note the odd placements of the markers on the lines).
I.1 Note that this is one of the few examples of a pure S v T exhibit.
Using one of Tufte's ideas, would the exhibit be improved by showing more than one 'cycle' e.g. 1½ days?
Question: what does the slope of each line represent, and what would 45° signify?
1.1 This is an interesting early example of small ‘sketches’ in a document - why don't authors do this more often?
1.4 Before the Cartesian paradigm: values shown in the plot; how would you revise?
1.5 Why break up by month?
What do the horizontal lines signify?
The grid is stronger than the plot.
1.6 Time labels on right
What of data between events?
Is an area chart appropriate?
1.7 The bars are ordered by "Today’s" population
I would show Z_t-38 above Z_t0 – we read down and left add % change.
P 23 Why does the book need foot- and endnotes?
T 4.1 This is a nicely designed table; I might use dittos or lines to signify the span of the monarchs.
4.1 How many dimensions of data?
What’s the extent & resolution of time?
Note that even 200 years ago ‘gradient’ was a temptation!
4.2 This datagraphic implies that measurements were made at the time events shown.
4.3 The gridlines (both styles) are distracting.
4.4 The time axis is mislabeled - should be the same as in 4.3.
The v-axis is too busy; (8, 16, 2) would be sufficient, without minor ticks.
4.5 Why are the dots connected by what appears to be a spline?
Would callouts (arrows + text boxes) be sufficient to indicate the events?
P 33 This is a nice graphical illustration of so-called 'spurious correlation.' The H-axis would be clearer with ticks and labels at 980 + 20 * (0, 6, 1). This is improved somewhat in 9.5.
4.6 What of the linear interpolation?
The contractions of the years on the H-axis isn't needed.
4.7 As with 4.2 a naive interpretation is that measurements were only made at the times of the labeled events - what of data at other times?
4.8 Note the labels on axis 4, where we can interpret current values more easily.
Why smooth the data?
4.9 What are the grid lines for?
5.2 This is a nice combination of T and S dimensions. 1¼ T cycles are shown. Although T is shown in the usual direction, because of Earth's sense of rotation S is therefore reversed.
5.3 A pure T graphic (except for the labels).
Study this graphic for awhile and realize that for T to increase to the right, events have to occur to the left! Think about this when making an animation.
6.1 This horizontal bar chart enables the V labels to be oriented as read.
The H labels could just as well be (0, 300, 20); note that the time scaling factor is implied in the last value.
The bar symbology may have been made on an early Macintosh computer; something subtler would do.
6.2 Another almost purely T datagraphic. The horizontal organization is always tricky: t_start, t_end, or (t_end - t_start)?
The labels are near the data.
It would be interesting to see a book containing the sketches of famous scientists!
7.3 This is very hard to read but rewards study.
Question What STP dimensions are shown here and on which axes?
8.1 Finally, we look at some maps!
How many variables do each of the symbol boxes illustrate?
Pressure could more clearly be shown with contours.
9.2 Using an ellipse to highlight a period is less clear than vertical line boundaries or a horizontal bold line.
The V-axis could be (0.5, 2, .05).
9.2 The log scale should be clearly labeled as such, and the minor ticks don't have much meaning (they're at 300, 3000, etc).
9.3 Why are there 3 font sizes on the T-axis, and what do the 5-year ticks mean?
9.5 A more sophisticated analysis would look at differences.
What is the smooth curve model based on?
15.3 QuestionWhat are at least 3 insights illustrated here (don't click on the link until you've thought about it!)?

EXHIBIT	COMMENTS
1,2	V-axis should be from 0 to 20 in steps of 5 - abbreviated (0, 20, 5). The minor ticks are unnecessary, especially in a non-time axis. As always in a time series, the V-axis annotations (title and tick labels) could be to the right. Show either 9 minor H-axis minor ticks (one for each year) or none at all. The exhibit could use a clear title instead of the V-axis label. Wainer's own graphics appear to be drawn by hand, or at least to have a slightly crude look (note the odd placements of the markers on the lines).
I.1	Note that this is one of the few examples of a pure S v T exhibit. Using one of Tufte's ideas, would the exhibit be improved by showing more than one 'cycle' e.g. 1½ days? Question: what does the slope of each line represent, and what would 45° signify?
1.1	This is an interesting early example of small ‘sketches’ in a document - why don't authors do this more often?
1.4	Before the Cartesian paradigm: values shown in the plot; how would you revise?
1.5	Why break up by month? What do the horizontal lines signify? The grid is stronger than the plot.
1.6	Time labels on right What of data between events? Is an area chart appropriate?
1.7	The bars are ordered by "Today’s" population I would show Z_t-38 above Z_t0 – we read down and left add % change.
P 23	Why does the book need foot- and endnotes?
T 4.1	This is a nicely designed table; I might use dittos or lines to signify the span of the monarchs.
4.1	How many dimensions of data? What’s the extent & resolution of time? Note that even 200 years ago ‘gradient’ was a temptation!
4.2	This datagraphic implies that measurements were made at the time events shown.
4.3	The gridlines (both styles) are distracting.
4.4	The time axis is mislabeled - should be the same as in 4.3. The v-axis is too busy; (8, 16, 2) would be sufficient, without minor ticks.
4.5	Why are the dots connected by what appears to be a spline? Would callouts (arrows + text boxes) be sufficient to indicate the events?
P 33	This is a nice graphical illustration of so-called 'spurious correlation.' The H-axis would be clearer with ticks and labels at 980 + 20 * (0, 6, 1). This is improved somewhat in 9.5.
4.6	What of the linear interpolation? The contractions of the years on the H-axis isn't needed.
4.7	As with 4.2 a naive interpretation is that measurements were only made at the times of the labeled events - what of data at other times?
4.8	Note the labels on axis 4, where we can interpret current values more easily. Why smooth the data?
4.9	What are the grid lines for?
5.2	This is a nice combination of T and S dimensions. 1¼ T cycles are shown. Although T is shown in the usual direction, because of Earth's sense of rotation S is therefore reversed.
5.3	A pure T graphic (except for the labels). Study this graphic for awhile and realize that for T to increase to the right, events have to occur to the left! Think about this when making an animation.
6.1	This horizontal bar chart enables the V labels to be oriented as read. The H labels could just as well be (0, 300, 20); note that the time scaling factor is implied in the last value. The bar symbology may have been made on an early Macintosh computer; something subtler would do.
6.2	Another almost purely T datagraphic. The horizontal organization is always tricky: t_start, t_end, or (t_end - t_start)? The labels are near the data. It would be interesting to see a book containing the sketches of famous scientists!
7.3	This is very hard to read but rewards study. Question What STP dimensions are shown here and on which axes?
8.1	Finally, we look at some maps! How many variables do each of the symbol boxes illustrate? Pressure could more clearly be shown with contours.
9.2	Using an ellipse to highlight a period is less clear than vertical line boundaries or a horizontal bold line. The V-axis could be (0.5, 2, .05).
9.2	The log scale should be clearly labeled as such, and the minor ticks don't have much meaning (they're at 300, 3000, etc).
9.3	Why are there 3 font sizes on the T-axis, and what do the 5-year ticks mean?
9.5	A more sophisticated analysis would look at differences. What is the smooth curve model based on?
15.3	QuestionWhat are at least 3 insights illustrated here (don't click on the link until you've thought about it!)?

2010/10/14