Sie sind auf Seite 1von 2

Graphics Principles Cheat Sheet v1.

1 Selecting the right base graph

Consider if a standard graph can be used by identifying suitable designs based on the:
Communication Principles of Effective Graphic Design (i) purpose (i.e. message to be conveyed or question to answer) and (ii) data (i.e. variables to display).

Effective visualizations communicate complex Example plots categorized by purpose

statistical and quantitative information facilitating Proximity – group related elements together
insight, understanding, and decision making. Deviation Correlation Ranking Distribution Evolution Part-to-whole Magnitude
Alignment – elements on the same vertical or
horizontal plane are perceived as having similar Chg. from Horizontal Kaplan Stacked Vertical
But what is an effective graph? Scatter plot Boxplot
properties baseline bar chart Meier bar chart bar chart
This cheat sheet provides general guidance and points Simplicity – cut anything superfluous, only
to consider. include elements that add value, limit to 2-3
colors or fonts

Planning White space (empty space) – use white space to

minimize distraction & provide clarity Waterfall Heat map Dotplot Histogram Line plot Tree map Forest plot

Clearly identify the purpose of the graph, Legibility – sans serif fonts are easier to read,
e.g. to deliver a message or for use color for emphasis instead of a new typeface
Why exploration? Color – select colors that present enough
contrast to make the graph legible. Choose
monochromatic color schemes to prevent
clashing. Use dark colors and accent colors to
Identify the quantitative evidence to emphasize important information Facilitating Comparisons Color for emphasis or distinction
support the purpose Visual Hierarchy – use color, font, image size,
What typeface, alignment & placement to create a Proximity improves association Restrained use of color is highly effective in organizing
viewing order a narrative and calling attention to certain elements.
Focal Points – primary area of interest that
Place labels next Think carefully before introducing additional color.
Identify the intended audience immediately attracts the eye, emphasize the most
to data instead of Do you really need it?
(specialists, non-specialists, both) and important concept and make it your focal point.
using legends
Who focus the design to support their needs Use contrasting colors to draw attention
Do not use color to
Repetition – repeating elements can be visually
appealing, repeated shapes, labels, colors differentiate between
Group together categories of the same
Familiarity – using familiar styles, icons, elements to be variable
Adapt the design to space or formatting
navigation structure makes viewers feel confident compared directly
constraints (e.g. clinical report, slide
Where deck or publication) Consistency – be consistent with heading sizes,
font choices, color scheme, and spacing. Use
Use colors or shades to
represent meaningful
images with similar styles Ease visual inspection differences such as
positive/negative values,
Order values to treatments or doses
Effectiveness Ranking help compare
across many Be consistent, use the
A graph is a representation of data that visually encodes numerical values into attributes such as lines, symbols and categories same color to mean the
colors. The Cleveland-McGill scale can be used to select the most effective attribute(s) for your purpose. same thing in a series of
graphs (e.g. treatment,
Position on Judgments are dose)
Depth: 3d Color Slope or Position on easier to make on
Volume Color hue Area Length unaligned
position intensity Angle common scale a common
vertical scale Use a bold, saturated or
contrasting color to
emphasize important
Reduce mental arithmetic details.

Plot the final

comparison e.g. Emphasize the data by
mean difference minimizing unnecessary
not two means ink, e.g. soften gridlines
Exception: if comparator
with a light color
Least accurate Most accurate is of interest in itself

bubble stacked bar dot plots, bar Use reference Utilize existing resources for selection of appropriate
volume multivariate charts, line graphs, charts, small charts, parallel lines and other palettes such as
designed heat maps
charts density plots mosaic pie charts waterfall multiple plots coordinate visual anchors. Color brewer
heat maps
charts chart plots or Munsell
Implementation Considerations Legibility and Clarity Good graph checklist
Plot cause on the x-axis Effective graphs stand alone. They use titles, Clear Communication Implementation Considerations
and effect on the y-axis. annotations, labels, shapes, colors, and textures to
Use this standard  Is the message of the graph as clear as  Are multiple panels plotted on the same
deliver important information. scale?
convention in order to possible?
avoid misinterpretation.
Label axes with clear  Is it easy for someone unfamiliar with the  Are lognormally distributed variables plotted
Aspect ratio can influence measurement units data to interpret the graph? on a log scale?
interpretation. Aim for a 45 and provide
degree angle of change to annotations that  Are the patterns/relationships easily  Are common baselines used wherever
avoid over-interpretation support the message. identified? possible?
of slope.
Use font size to create  Is the graph tailored to its primary purpose  Does the orientation of the axes aid
Use position for hierarchy (e.g. set titles and audience? interpretation?
comparisons rather than 2pt larger than all other
length (i.e. dots instead of labels to make them  Is the correct graph type used?  Does the aspect ratio allow the reader to
bars), especially for non- more prominent). see variations in the data?
linear scales (e.g. log
scale or % change). Facilitating Comparisons  Are data across a disconnected time scale
Do not type too small or kept disconnected?
Do not plot log-normally too condensed. Break
 Are elements to be compared grouped
distributed variables on a long titles into two lines.  Are data spaced proportionally to the actual
linear scale (e.g. hazard Shift or adjust size of time interval (instead of according to visit
ratio, AUC, CL) labels that overlap.  Are labels placed next to data instead of in number)?
 Are data and inferences plotted to support
When displaying data Keep the font style  Have categories been ordered for easy stories about models?
measured on the same simple – sans serif is
scale, also plot them on comparison?
easier to read.  Are number of patients by group reported if
the same scale for easy
 Can the plot be read without doing mental this adds context?
Display text with enough
Connected data imply contrast to be visible.  Are the estimates of interest plotted (e.g. Legibility and Clarity
continuity. Do not connect Favor the use of dark on mean differences with confidence
light instead of light on  Can all graphical elements be seen?
data across a disconnected intervals)?
or uneven time scale. dark whenever possible.
 Does the graph have a clear title, axis
Visits displayed close labels, annotations and data units?
Bold or italics should only Color for emphasis or distinction
together are perceived to be used for layering or  Can the font be read without eye strain or
be closer in time. Space emphasis. Emphasizing  Are graphical elements displayed in a dark effort?
the visits proportional to everything means nothing color on a light background?
the time between each in gets emphasized.  Are sans-serif fonts used?
order to avoid confusion.  Are grid lines drawn with a thin line and a
Exception: baseline or pre-dose
Try not to set text at an light color such as grey?  Do text sizes have correct hierarchy (big to
angle, as this decreases small, main text to subtext)?
Plot data and inferences  Are colors used sparingly (e.g. max 3)?
readability. Think of
to support stories about  Are the elements of the graph clearly
alternative solutions such Do all elements in the graph have a purpose
models.  labeled (e.g. points, error bars, lines,
as transposing the graph.
(e.g. colors, textures, grid lines)? shaded regions)?
 Are the same colors used to mean the same  Are labels oriented horizontally where
Putting it all together – Remove the clutter & emphasize the message thing in a series of graphs? possible?
Creating a graph is an iterative process: produce, review and refine.

Books: Online resources:
E. R. Tufte, The visual display of quantitative information, Connecticut, Graphics. Press, 2001. (S. Few)
Cleveland, W.S. and McGill, Robert, Graphical perception: theory, experimentation and application
to the development of graphical methods, JASA, Vol. 79, No. 387, pp. 531 – 554, 1984. (E. Tufte)
S. Few, Show Me The Numbers ‐ Designing Tables and Graphs to Enlighten (2nd. Edition), (N. Yau)
Burlingame, CA: Analytics Press, 2012. (J. Doumont)
Colors, It is easier to see Using too many colors can One solution could be D. M. Wong, The Wall Street Journal Guide to Information Graphics: The Dos and Don'ts of (A. Gelman)
Presenting Data, Facts, and Figures. December 16, 2013. (A. Cairo)
backgrounds, and differences in be distracting. Use white repeating the data in
J. Doumont, Trees, maps, and theorems: Effective communication for rational minds. PRINCIPIAE. (N. Robbins)
borders can be position over a background and try using different panels, N. B. Robbins, Creating More Effective Graphs. Chart House.
removed and difference in length, other methods to highlighting individual
gridlines reduced. i.e. a dot over a bar. distinguish different curves. curves in a darker color.
Alison Margolskee, Mark Baillie, Baldur Magnusson, Julie Jones, Marc Vandemeulebroecke
© 2017 Novartis