Statistical results, poorly
communicated, will not have an impact on the problem or the intended
audience. You should craft your communications to meet the needs of
your audience in the time or space allowed. Do not try to force 10
minutes of information into a 5 minute presentation. In some cases
you will need to prepare different communications for different audiences
(e.g., general public, scientific or professional community), taking
into consideration their familiarity with the problem domain, terminology,
and statistical analysis. For example, in conveying statistical results
to the general public it may be sufficient to state that a result
is statistically significant, while communicating the same result
to a scientific audience may benefit by reporting a p-value.
The following guidelines can assist you in preparing
effective data analysis presentations:
-
Focus on the problem and how your
analysis adds insight.
-
Use the language of the problem
domain.
-
Avoid the use of statistical jargon
and notation.
-
Organize your information in a
way that makes sense to your audience.
-
Familiarize your audience with
the data.
-
Identify the data source and give
data definitions for key variables.
-
Numerical summaries are easily
assimilated in tables. Round numbers appropriately and display units.
-
Visualizations allow data distributions,
outliers, time trends, and geographic relationships to be easily perceived.
Include titles, axis scales and labels. Use color judiciously.
-
Focus on the essential analytic
results as related to the problem
-
Always identify results as being
statistically significant or insignificant.
-
State the limitations of the analysis.
-
Identify insights gained from the
analysis and the effect of unusual observations.
Your information will
be best received if it is focused, concise, relevant, and delivered
with enthusiasm.
The twelve cases presented
in this book illustrate many of the concepts covered in this introduction.
With the exception of one (Visualizing Influenza Activity), the cases
focus on six different problems with several cases forming a sequence
for each problem. This illustrates a problem-solving approach starting
with basic descriptive analysis and building to more complicated statistical
inference and modeling. All but one data set were obtained from open
sources.