A table of bar graphs (small multiples)

“	Power corrupts. PowerPoint corrupts absolutely.” — Edward R. Tufte

With the greatest respect to Professor Tufte, whom I consider my teacher,
I propose the design below as an improvement on the design
from his monograph, The Cognitive Style of PowerPoint,
summarized in Wired September 2003, pp 118-119.

Estimated survival rates by cancer site

% survival / standard error

5 year

10 year

15 year

20 year

Prostate

98.8

0.4

95.2

0.9

87.1

1.7

81.3

3.0

Thyroid

96.0

0.8

95.8

1.2

94.0

1.6

95.4

2.1

Testis

94.7

1.1

94.0

1.3

91.1

1.8

88.2

2.3

Melanomas

89.0

0.8

86.7

1.1

83.5

1.5

82.8

1.9

Breast

86.4

0.4

78.3

0.6

71.3

0.7

65.0

0.7

Hodgkin’s Disease

85.1

1.7

79.8

2.0

73.8

2.4

67.1

2.8

Corpus uteri, uterus

84.3

1.0

83.2

1.3

80.8

1.7

79.2

2.0

Urinary, bladder

82.1

1.0

76.2

1.4

70.3

1.9

67.9

2.4

Cervis, uteri

70.5

1.6

64.1

1.8

62.8

2.1

60.0

2.4

Larynx

68.8

2.1

56.7

2.5

45.8

2.8

37.8

3.1

Rectum

62.6

1.2

55.2

1.4

51.8

1.8

49.2

2.3

Kidney, renal pelvis

61.8

1.3

54.4

1.6

49.8

2.0

47.3

2.6

Colon

61.7

0.8

55.4

1.0

53.9

1.2

52.3

1.6

Non-Hodgkin’s

57.8

1.0

46.3

1.2

38.3

1.4

34.3

1.7

Oral cavity, pharynx

56.7

1.3

44.2

1.4

37.5

1.6

33.0

1.8

Ovary

55.0

1.3

49.3

1.6

49.9

1.9

49.6

2.4

Leukemia

42.5

1.2

32.4

1.3

29.7

1.5

26.2

1.7

Brain, nervous system

32.0

1.4

29.2

1.5

27.6

1.6

26.1

1.9

Multiple myeloma

29.5

1.6

12.7

1.5

7.0

1.3

4.8

1.5

Stomach

23.8

1.3

19.4

1.4

19.0

1.7

14.9

1.9

Lung and bronchus

15.0

0.4

10.6

0.4

8.1

0.4

6.5

0.4

Esophagus

14.2

1.4

7.9

1.3

7.7

1.6

5.4

2.0

Liver, bile duct

7.5

1.1

5.8

1.2

6.3

1.5

7.6

2.0

Pancreas

4.0

0.5

3.0

1.5

2.7

0.6

2.7

0.8

A table of bar graphs (small multiples): rich, informative, visually comparative.

I think I prefer an alternate version with drop shadow and white instead of dark gray,
though it doesn’t work in Netscape (through 7.1). (Thanks to Jeffrey Berg for
prodding me to work this out.)

A further improvement would depict the standard error as a dimmed zone between the two percentage bars, perhaps something like this:

Or perhaps the border between the two colors could have a slant, the more slant, the more standard error. But first I need a function that takes percent and standard error as input and returns the appropriate 4 percentages to display...

Discuss this on Edward Tufte’s web site.

(If the small type looks bad on your computer, try a Mac!)