Design of Scientific Graphics

1
Design of Scientific Graphics
Prof. Melissa A. Hines
2
Design of Scientific Graphics for Publication

• Prepare all graphics at full scale (see journal
  rules).

Note that journals often compress graphics w/o
asking.
3
Design of Scientific Graphics for Publication

• General rules

• Use bland san serif font (typ Helvetica)
  throughoutRegular weight, not bold
• Major elements in graph (e.g., axes) 1 pt
  minMinimum linewidth of any feature 0.5 pt
• Check journal rules on color before starting
• Some letters journals limit number of graphics

Choose designs that make your graphs
self-explanatory. Dont use captions as a crutch!
4
Typography Gotchas for Scientific Graphics

• Italicize all variables in text and in graphics
  x-axis not x-axis, f not f (
  but µm, K, C, etc. are not italicized)
• Use the appropriate length dashes - (hyphen)
  Use to hyphenate words (en dash) minus
  sign (em dash) Use for punctuating sentences
• Use the correct symbol for multiplication
  ? not x, (? 103 m) not (x 103 m)

5
Legal Gotchas for Scientific Graphics

• Publisher owns all images in a published paper
  (but not underlying data).
• Cannot reuse figures from one paper to another
  without attribution and publisher's written
  consent.
• Trick Modifying original figure makes it "new"

Note You do not need permission to use someone
else's data (taken from a published source).
6
Is the Golden Ratio Golden?
1.618
1
Choose the aspect ratio that flatters your
data, not some dead Greek philosopher.
7
To Box or Not To Box?

• Vs.

Pro Minimal design
Pro Easier to judge scale on RHS of graph
8
Should Zero Be Included?

• Vs.

9
Graphics Affect Perception of Data
Published graph suggests a correlation
10
Small Signals and Baseline Issues
Baseline must be shown on both sides of peaks.
11
Appropriate Use of the Top or Right Axis
Use second axis to show data in different units.
12
Less Appropriate Use of the Top or Right Axis

• Plotting different types of data on opposite axes
  is outdated.
• In choosing software, preparation of stacked
  graphs (e.g., same bottom axis) is important.

13
Use Labels to Remove Dependence on Captions
Column and row headers explain different parts of
graph w/o captions. Individual labels on
sub-images only necessary for reference in text.
(Do not include in presentations!) Note all
images have same scale bar.
14
Use Shading to Convey Temporal Information
Shading is particularly useful in indicating
changes in experimental conditions.
15
Use Shading to Distinguish Curves
The histograms (data) are shaded for emphasis and
to help the reader perceive their shape. The
limiting curves (exp Gaussian) are left
unfilled for reference.
16
Use Shading to Distinguish Curves
The shaded curve represents historical data the
black line is the new data. This emphasizes that
the two spectra have different peaks
different linewidths
17
Improving Contrast of Overlapping Lines
Problem New data (black) doesn't stand out from
old data (red shading)
18
Focusing Attention on Specific Regions
Differential absorption spectrum
Common Solution Plot regions of interest,
leaving breaks in axes for "boring"
areas. Caution Scale should remain constant
across all regions!
19
Focusing Attention on Specific Regions
Another Solution Use insets to plot regions of
interest.
20
Legends should reinforce data where possible
If possible, put legend in the same order as
items on the graph. Jumbled order will confuse
the reader.
21
Comparing Similar Data with Residuals
Problem Nearly identical curves cannot be
distinguished when plotted on same graph. Offset?
22
Curve Fitting and Residuals
A structured residual is indicative
ofinappropriate modeling function.
23
What about pie charts?
Pie charts have no place in science. Well-documen
ted problems with accurate perception of relative
areas.
24
Minimum Needs for Scientific Graphing Package

• Support for equally spaced x-y data
• Ability to easily transform data (e.g., multiply,
  divide, FFT)
• Curve fitting to user-defined functions
  user-defined regions of data. Global fits to
  multiple data sets.
• Control over all parts of graph (e.g., ticks,
  axes, labels)
• Multiple, arbitrarily positioned and scaled axes
  (e.g., stacked graphs, insets)
• Robust data import graphics export
• Fully documented analysis functions w/
  references
• Active responsive user network for help

25
Your Assignment Help DOE Communicate!
This graph has no discernable message! Make a
new publication-quality graph that emphasizes
Contributions to US energy landscape by energy
type vs year Percentage of renewable energy vs
year
26
Your Tool Igor Pro
You can install Igor Pro on your personal
computer or a lab computer for use in completing
the assignment. You cannot use the program for
research or personal use without buying a license.