Fabrication, Falsification, and the Sanctity of Data

1
Fabrication, Falsification, and the Sanctity of
Data

• Prof. William Ullman
• College of Marine and Earth Studies
• University of Delaware, Lewes

2
Federal Policy on Research Misconduct
Research misconduct is defined as fabrication,
falsification, or plagiarism in proposing,
performing, or reviewing research, or in
reporting research results.
US Office of Science and Technology Policy.
ltwww.ostp.gov/html/001207_3.htmlgt
3
Fabrication

• Fabrication is the description of experiments not
  actually performed, the invention of data not
  actually collected, and/or the reporting of these
  experiments and results.

4
Falsification (Cooking and Trimming)

• Falsification is manipulating research materials,
  equipment, or processes, or changing or omitting
  data or results such that the research is not
  accurately represented in the research record.
• Cooking is retaining and reporting only the data
  that fits the theory and discarding others.
• Trimming is the smoothing of irregularities to
  make the data look more accurate and precise than
  they really are.

5
Nullius in Verba!

• Colleagues/readers are entitled to

• See all of the necessary data
• Know how the data was collected
• Know the limits of the methods and
• Make their own judgments based on your data!

Honor and Integrity in Science
6
Deborahs and Kathleens DataA Case Study

• Deborah (graduate student) and Kathleen (post
  doc) make expensive measurements at a national
  laboratory to verify a newly proposed theory.
  When they complete the experiment and return to
  their own lab, they review their data and compare
  it with the new theory.

From On Being a Scientist Responsible
Conduct in Research, 2nd Edition. National
Academy Press, 1995
7
Prediction from Theory
8
Deborahs and Kathleens DataA Case Study

• During the experiments at the national
  laboratory, they observed unpredictable,
  uncontrollable, and unexplained fluctuations in
  the responses for two data points that fell the
  furthest from the theoretical prediction. They
  also found out that another research group,
  pursuing similar experiments, had independently
  verified the proposed theory.

9
Deborahs and Kathleens DataA Case Study

• Kathleen suggests that, due to the observed
  fluctuations, these points be omitted from the
  statistical analysis, but, of course, be reported
  in the paper to be published from this
  experiment. In the text they will say that these
  points reflect anomalies associated with power
  fluctuations and are outside of the uncertainty
  associated with all of the points.

10
Deborahs and Kathleens DataA Case Study

• How should the data from the two points be
  handled?
• Should the data be included in the statistical
  tests?
• Who can they go to for advice?
• Is this paper ready to be written?

11
Deborahs and Kathleens DataA Case Study

• Are there problems with Deborahs and Kathleens
  approach to their data?
• How would you examine this data?
• What is a datum?
• Is there something
• called self-deception?

12
(No Transcript)
13
(No Transcript)
14
(No Transcript)
15
(No Transcript)
16
(No Transcript)
17
Report all as measured without uncertainty
Cooking
Trimming
Report all as measured with uncertainty
Eliminate on a posteriori analysis
Eliminate on deviation from expectations
Eliminate experimental anomalies
Fabricate experiment data