Professional Documents
Culture Documents
Decisions Made
Graphical Excellence
Have the audience in mind. What is the purpose of the graphic?
Description, exploration
Performane Rate
0.8
0.6
1) Maximize data-ink ratio. 0.4
0.2
I. Erase non-data-ink 0
1 2 3 4 5 6 7 8 9
II. Erase redundant Clinic
data-ink
After
2) Remove Chart Junk. Clinical Visits
I. Shadows Percentage of adult patients who had at least one
visit in each half of the year
II. 3D-rendering 1
Performane Rate
0.8
III. Other ornaments 0.6
0.4
Tufte, Edward. The Visual Display of Quantitative Information. Connecticut, Graphic Press: 2001. Page 13.
120
100
80
60
40
20
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19
Examples
Bar Charts
Good for comparing a set of categorical values. Best when there are not too
many categories and/or variables.
Clinical Visits
Percentage of adult patients who had at least one visit in
each half of the year
1
0.8
Performane Rate
0.6
0.4
0.2
0
1 2 3 4 5 6 7 8 9
Clinic
Tips:
Organizing data from largest to smallest may be helpful in highlighting data.
Keep it simple: do not use shadows or 3D rectangles.
Too many categories can make bar charts messy. When there are this many bars on a bar
graph, make sure to ask yourself if it is contextually appropriate to compare all of the
values on the bar chart.
70
60
50
40
30
20
10
0
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
Clinic
Too many variables per category can also make bar charts messy. Is it appropriate to
compare all of the variables within a category?
70
Clinical Visits
60
TB Screening
50
CTX
40
Nutritional Assessment
30 Prevention Education
20 Alcohol Screening
10
0
A B C D E
Clinic
Pie Charts
Work well if you want to compare individual slices of the pie with the whole
pie. It may be difficult to compare different sections of a given pie chart or
to compare data across different pie charts. A bar chart (histogram or stack
chart) or table may be more appropriate in that case.
Too many variables make a pie chart hard to manage. If the variables are
numerical, consider using a histogram instead. You can also consider combining
categories but remember that this could hide variation and alter how the data
are interpreted.
<50
CD4 Count Distribution 51-100
101-150
CD4 Count Distribution
151-200
201-250
251-300
301-350
351-400 <50
451-500
501-550 51-100
551-600
601-650 101-200
651-700 201-250
701-750
751-800 400+
801-850
851-900
901-950
951-1000
1000+
Tables
Tables often work better than bar charts and pie charts when there are too many data
points and too many descriptors of those data points. Many people may not consider
this as a way to visualize data, but tables still use specific formatting and spatial
orientation to communicate the data more easily. In terms of data ink, every piece of
a table is critical information. However, tables may not be good at showing patterns
over time.
CD4 Monitoring Mean Clinic Scores
Percentage of eligible patients who had at least one CD4
count during the review period
Table Formatting Tips
Do not use gridlines. The space between the numbers visually separate categories.
Underline the column headers
Consider Zebra Striping: light shading to separate specific groups you want to
highlight.
Tips:
Use different colors to differentiate between different line. Remember that our eyes
will naturally compare two different lines on the same chart. If two data points are not
comparable, then maybe they should not be on the same graph.
Label the lines directly on the chart instead of using a legend.
Line charts are very prone to distortion.
Percentage of eligible patients screened for tuberculosis
Y Axis Scale: 0 to 25 Y Axis Scale: 0 to 100
25 100
20
Performance Rate (%)
15
50
10
25
5
0 0
Jan Feb Mar Apr May June Jan Feb Mar Apr May June
25 Y Axis Scale: 15 to 20
20
20
Performance Rate (%)
Y Axis Scale: 0 to 25
Performance Rate (%)
19
15
Height > Width 18
10 17
16
5
15
Jan Feb Mar Apr May June
0
Mar
Jan
Apr
June
May
Feb
Box-and-whisker Plots
Are a great way to compare different sets of data. Several
different descriptive statistics can be compared: Max, min,
upper quartile, median, lower quartile, range and interquartile
range.
Namibia Food Security
Oct 10 - Mar 11
Jan - Jun 10
Jul - Dec 09
Review Period
Jan - Jun 09
Jul - Dec 08
Jan - Jun 08
0 10 20 30 40 50 60 70 80 90 100
Performance Rate (%)
The next few examples illustrate how important labeling is.
Labeling provides more context to the data, allowing for
more rigorous and accurate interpretations of the data.
12
10
people/year)
0
Soccer Rugby Cricket Golf
0
2
4
6
8
10
12
Average Age = 23
Soccer
Average Age = 20
Rugby
in 2011
Average Age = 25
Cricket
Average Age = 60
Golf
Mortality Rate of People Actively Playing Popular Sports
What can we conclude?
Percent of Adults who received a TB assessment
during the review period (Adult, 2008)
100
90
80
Performance Rate (%)
70
60
50
40
30
20
10
0
Clinic A Clinic B Clinic C
Percent of Adults who received a TB assessment
during the review period (Adult, 2008) n = 2
100
90
n = 150
80 n = 200
Performance Rate (%)
70
60
50
40
30
20
10
0
Clinic A Clinic B Clinic C
Jul Dec
Jan Jun
Jan Jun
Jan Jun
Heat Maps
Use color to encourage the eye to
examine both clinic level and aggregate A
level patterns. In this example, each B
color represents a range of performance
rates. The more red the color, the closer C
the performance rate is to 0%. The D
more green the color the closer the E
performance rate is to 100%.
F
G
Clinic
Namibia Food Security H
Key to Swatch
Indicator Results I Colors
Percentage of eligible adult patients J Rate (%)
0 to 10
assessed for food security by clinic
K 11 to 20
and review period. 21 to 30
L 31 to 40
41 to 50
M 51 to 60
N 61 to 70
71 to 80
O 81 to 90
91 to 100
P
Summary
Context is essential for graphical integrity.
Provide historical data when available.
Label axes properly.
Always provide denominators to percentages.