Interpreting Scientific
Figures and Statistics
The majority of rare diseases begin
childhood. Approximately 30% of
children with rare diseases will not
live to their 5th birthday.
Understanding and interpreting scientific figures is a critical skill for scientists. Putting data into a visual
format helps scientists analyze, understand, and share their results; however, data visualizations, or the
graphic representation of data, can also sometimes be confusing or misleading. With all the data shared
on the news and social media which illustrates important issues like disease outbreaks and genetic risks,
the ability to critically interpret these data visualizations has become more important than ever. If you
look out for a few common errors, you can be a savvier consumer of data visualizations.
Common Error 1: Misleading Axes
Always check the axes of graphs because different axes formats can greatly affect how the data
in a graph look. The horizontal or X-axis reads from left to right and the vertical or Y-axis reads
bottom to top. For both axes, the scale, or interval between ticks, (e.g., 0-1500 on the left Y-axis
below) and the labeled increments (e.g., 0, 500, 1000, 1500 on the left Y-axis below) should be
suitable for the data. The graphs below demonstrate how changing the axis scale and increments
can make a big difference. Both graphs show the number of essays submitted to the ASHG DNA
Day Essay Contest from 2015 to 2021; however, different scales and increments are used. In the
left graph, a linear scale is used, and each increment represents 500 essays. In the right graph, a
logarithmic scale is used where each y-tick represents an increasing factor of 10. While logarithmic
scales can be useful for visualizing data with a very large spread, it is misleading in this case
because it makes it look like the number of submissions stayed about the same each year while
the linear graph on the left shows that the number of submissions more than doubled from
2015 to 2016.
6120 Executive Blvd, Suite 500, Rockville, Maryland 20852 · (301) 634-7300 · www.ashg.org
@GeneticsSociety
Source: ASHG
Common Error 2: Confusing Correlation with Causation
When two variables are correlated, it means that there is a pattern in the data. So, as one variable changes, the
other variable also changes. This may make it seem like the change in one variable is causing the change in the
other, but that is not always the case. A strong correlation could indicate causality, but there can be other
explanations. In some cases, changes in a third variable could be causing similar changes in the variables shown
in the graph. For example, as the temperature increases, the amount of ice cream sold and the number of people
with sunburns at the beach will both increase. If you graphed ice cream sales and sunburn on the same graph,
some might assume that the increase in one is causing the increase in the other; however, it is really the increase in
temperature that is causing the change in the other two variables.
6120 Executive Blvd, Suite 500, Rockville, Maryland 20852 · (301) 634-7300 · www.ashg.org
@GeneticsSociety
A correlation between two variables could also be due to random chance, where the variables appear to be
related but there is no true underlying relationship between them. The graph below plots the amount of money
spent on pets (green data) against the number of lawyers in California (blue data) between 2000 and 2009.
Both appear to have increased at the same rate over time; however, it is unlikely that either increase caused the
other increase.
Money spent on pets vs. the number of lawyers in California from 2000 to 2009
Source: Spurious Correlations
More Questions to Ask
• Do the visuals (e.g., graph, pictogram) match the numbers? For example, do the sizes of pieces in a pie
graph match the percentages given?
• When and how was the data collected? For example, consider data collected about what type of music
Americans listen to. You can imagine that the data collected on a college campus may be different from data
collected outside an opera hall. Make sure that when and how the data was collected will answer the question
being asked.
• Who funded the study? Who else is involved in getting the message out? Someone with a vested interest in
the data could manipulate the data presentation.
