Ethical Guidelines

for Statistical Practice

Prepared by the Committee
on Professional Ethics of the
American Statistical Association

Approved by the ASA Board in April 2016

Contents
Purpose of the Guidelines 1
A. Professional Integrity and
Accountability 2
B. Integrity of Data and Methods 3
C. Responsibilities to Science/Public/
Funder/Client 4
D. Responsibilities to Research
Subjects 5
E. Responsibilities to Research Team
Colleagues 6
F. Responsibilities to Other
Statisticians or Statistics
Practitioners 7
G. Responsibilities Regarding
Allegations of Misconduct 8
H. Responsibilities of Employers,
Including Organizations,
Individuals, Attorneys, or Other
Clients Employing Statistical
Practitioners 9
Discussion 10
Purpose of the Guidelines
The American Statistical Association’s of advancing knowledge while avoiding
Ethical Guidelines for Statistical Practice harm; using statistics in pursuit of unethi-
are intended to help statistics practitioners cal ends is inherently unethical.2
make decisions ethically. Additionally, the
Ethical Guidelines aim to promote ac-
countability by informing those who rely Good statistical
on statistical analysis of the standards that
they should expect. The discipline of statis- practice is
tics links the capacity to observe with the fundamentally
ability to gather evidence and make deci-
sions, providing a foundation for building based on
a more informed society. Because society transparent
depends on informed judgments supported
by statistical methods, all practitioners of
assumptions,
statistics, regardless of training and occupa- reproducible
tion or job title, have an obligation to work
in a professional, competent, and ethical
results, and valid
manner and to discourage any type of pro- interpretations.
fessional and scientific misconduct.1

Good statistical practice is fundamentally The principles expressed here should guide
based on transparent assumptions, repro- both those whose primary occupation is
ducible results, and valid interpretations. statistics and those in all other disciplines
In some situations, Guideline principles who use statistical methods in their profes-
may conflict, requiring individuals to sional work. Therefore, throughout these
prioritize principles according to context. Guidelines, the term “statistician” includes
However, in all cases, stakeholders have all practitioners of statistics and quantita-
an obligation to act in good faith, to act tive sciences, regardless of job title or field
in a manner that is consistent with these of degree, comprising statisticians at all
Guidelines, and to encourage others to levels of the profession and members of
do the same. Above all, professionalism other professions who utilize and report
in statistical practice presumes the goal statistical analyses and their implications.

1 Ethical Guidelines for Statistical Practice

American Statistical Association
A. Professional Integrity
and Accountability
The ethical statistician uses methodology beyond the sample to a population
and data that are relevant and appropri- relevant to the objectives with minimal
ate, without favoritism or prejudice, and error under reasonable assumptions.5
in a manner intended to produce valid, 3. Respects and acknowledges the con-
interpretable, and reproducible results.3 tributions and intellectual property
The ethical statistician does not knowingly of others.
accept work for which he/she is not suffi-
ciently qualified, is honest with the client 4. When establishing authorship order for
about any limitation of expertise, and posters, papers, and other scholarship,
consults other statisticians when necessary strives to make clear the basis for this
or in doubt. order, if determined on grounds other
than intellectual contribution.6
The ethical statistician: 5. Discloses conflicts of interest, financial
and otherwise, and manages or resolves
1. Identifies and mitigates any preferences them according to established (institu-
on the part of the investigators or data tional/regional/local) rules and laws.7
providers that might predetermine or 6. Accepts full responsibility for his/her
influence the analyses/results.4 professional performance. Provides
2. Employs selection or sampling meth- only expert testimony, written work,
ods and analytic approaches appropri- and oral presentations that he/she
ate and valid for the specific question would be willing to have peer reviewed.
to be addressed, so that results extend

2 Ethical Guidelines for Statistical Practice

American Statistical Association
B. Integrity of Data and Methods
The ethical statistician is candid about any 6. Where appropriate, addresses potential
known or suspected limitations, defects, or confounding variables not included in
biases in the data that may impact the in- the study.
tegrity or reliability of the statistical analy- 7. In publications and reports, conveys
sis. Objective and valid interpretation of the the findings in ways that are both
results requires that the underlying analysis honest and meaningful to the user/
recognizes and acknowledges the degree of reader. This includes tables, models,
reliability and integrity of the data.8 and graphics.
The ethical statistician: 8. In publications or testimony, identi-
fies the ultimate financial sponsor of
1. Acknowledges statistical and substan- the study, the stated purpose, and the
tive assumptions made in the execution intended use of the study results.
and interpretation of any analysis. 9. When reporting analyses of volunteer
When reporting on the validity of data data or other data that may not be
used, acknowledges data editing pro- representative of a defined population,
cedures, including any imputation and includes appropriate disclaimers and, if
missing data mechanisms. used, appropriate weighting.
2. Reports the limitations of statistical in- 10. To aid peer review and replication,
ference and possible sources of error.9 shares the data used in the analyses
3. In publications, reports, or testimony, whenever possible/allowable, and exer-
identifies who is responsible for the sta- cises due caution to protect proprietary
tistical work if it would not otherwise and confidential data, including all
be apparent. data that might inappropriately reveal
respondent identities.
4. Reports the sources and assessed ade-
quacy of the data; accounts for all data 11. Strives to promptly correct any errors
considered in a study and explains the discovered while producing the final
sample(s) actually used. report or after publication. As appro-
priate, disseminates the correction pub-
5. Clearly and fully reports the steps licly or to others relying on the results.
taken to preserve data integrity and
valid results.

3 Ethical Guidelines for Statistical Practice

American Statistical Association
C. Responsibilities to
Science/Public/Funder/Client
The ethical statistician supports valid in- 3. Applies statistical sampling and anal-
ferences, transparency, and good science in ysis procedures scientifically, without
general, keeping the interests of the public, predetermining the outcome.
funder, client, or customer in mind (as well 4. Strives to make new statistical knowl-
as professional colleagues, patients, the edge widely available to provide bene-
public, and the scientific community).1⁰ fits to society at large and beyond his/
her own scope of applications.11
The ethical statistician:
5. Understands and conforms to confi-
1. To the extent possible, presents a client dentiality requirements of data collec-
or employer with choices among valid tion, release, and dissemination and
alternative statistical approaches that any restrictions on its use established
may vary in scope, cost, or precision. by the data provider (to the extent
legally required), and protects use and
2. Strives to explain any expected ad- disclosure of data accordingly. Guards
verse consequences of failure to follow privileged information of the employer,
through on an agreed-upon sampling client, or funder.
or analytic plan.

4 Ethical Guidelines for Statistical Practice

American Statistical Association
D. Responsibilities to
Research Subjects
The ethical statistician protects and re- 4. Knows the legal limitations on priva-
spects the rights and interests of human cy and confidentiality assurances and
and animal subjects at all stages of their does not over-promise or assume legal
involvement in a project. This includes privacy and confidentiality protections
respondents to the census or to surveys, where they may not apply.
those whose data are contained in adminis- 5. Considers whether appropriate re-
trative records, and subjects of physically or search-subject approvals were ob-
psychologically invasive research. tained before participating in a study
involving human beings or organiza-
The ethical statistician: tions, before analyzing data from such
a study, and while reviewing manu-
1. Keeps informed about and adheres to scripts for publication or internal use.
applicable rules, approvals, and guide- The statistician considers the treat-
lines for the protection and welfare of ment of research subjects (e.g., con-
human and animal subjects.12 fidentiality agreements, expectations
2. Strives to avoid the use of excessive of privacy, notification, consent, etc.)
or inadequate numbers of research when evaluating the appropriateness
subjects, and excessive risk to research of the data source(s).14
subjects (in terms of health, welfare, 6. In contemplating whether to partic-
privacy, and ownership of their own ipate in an analysis of data from a
data), by making informed recommen- particular source, refuses to do so if
dations for study size.13 participating in the analysis could rea-
3. Protects the privacy and confidential- sonably be interpreted by individuals
ity of research subjects and data con- who provided information as sanction-
cerning them, whether obtained from ing a violation of their rights.
the subjects directly, other persons, or 7. Recognizes that any statistical descrip-
existing records. Anticipates and solic- tions of groups may carry risks of ste-
its approval for secondary and indirect reotypes and stigmatization. Statisticians
uses of the data, including linkage should contemplate, and be sensitive
to other data sets, when obtaining to, the manner in which information is
approvals from research subjects, and framed so as to avoid disproportionate
obtains approvals appropriate to allow harms to vulnerable groups.
for peer review and independent repli-
cation of analyses.

5 Ethical Guidelines for Statistical Practice

American Statistical Association
E. Responsibilities to
Research Team Colleagues
Science and statistical practice are often do not have obligations to standards
conducted in teams made up of profession- of other professions that conflict with
als with different professional standards. these Guidelines.
The statistician must know how to work 2. Ensures that all discussion and report-
ethically in this environment. ing of statistical design and analysis is
consistent with these Guidelines.
The ethical statistician:
3. Avoids compromising scientific validity
1. Recognizes that other professions have for expediency.15
standards and obligations, that research 4. Strives to promote transparency in
practices and standards can differ design, execution, and reporting or
across disciplines, and that statisticians presenting of all analyses.

6 Ethical Guidelines for Statistical Practice

American Statistical Association
F. Responsibilities to Other
Statisticians or Statistics Practitioners
The practice of statistics requires consid- 2. Helps strengthen the work of others
eration of the entire range of possible ex- through appropriate peer review; in
planations for observed phenomena, and peer review, respects differences of
distinct observers drawing on their own opinion and assesses methods, not
unique sets of experiences can arrive at individuals. Strives to complete review
different and potentially diverging judg- assignments thoroughly, thoughtfully,
ments about the plausibility of different and promptly.
explanations. Even in adversarial settings, 3. Instills in students and non-statisticians
discourse tends to be most successful an appreciation for the practical value
when statisticians treat one another with of the concepts and methods they are
mutual respect and focus on scientific learning or using.
principles, methodology and the sub-
stance of data interpretations. 4. Uses professional qualifications and
contributions as the basis for decisions
Out of respect for fellow statistical practi- regarding statistical practitioners’
tioners, the ethical statistician: hiring, firing, promotion, work assign-
ments, publications and presentations,
1. Promotes sharing of data and methods candidacy for offices and awards, fund-
as much as possible and as appropriate ing or approval of research, and other
without compromising propriety. Makes professional matters.
documentation suitable for replicate 5. Does not harass or discriminate.
analyses, metadata studies, and other
research by qualified investigators.

7 Ethical Guidelines for Statistical Practice

American Statistical Association
G. Responsibilities Regarding
Allegations of Misconduct
The ethical statistician understands the 4. Maintains confidentiality during an
difference between questionable scientific investigation, but discloses the in-
practices and practices that constitute mis- vestigation results honestly to appro-
conduct, avoids both, but knows how each priate parties and stakeholders once
should be handled. they are available.
5. Following an investigation of miscon-
The ethical statistician: duct, supports the appropriate efforts of
all involved, including those reporting
1. Avoids condoning or appearing to con- the possible scientific error or mis-
done incompetent or unethical practic- conduct, to resume their careers in as
es in statistical analysis. normal a manner as possible.
2. Recognizes that differences of opinion 6. Avoids, and acts to discourage,
and honest error do not constitute mis- retaliation against or damage to the
conduct; they warrant discussion, but employability of those who responsi-
not accusation. bly call attention to possible scientific
3. Knows the definitions of, and pro- error or misconduct.
cedures relating to, misconduct. If
involved in a misconduct investigation,
follows prescribed procedures.

8 Ethical Guidelines for Statistical Practice

American Statistical Association
H. Responsibilities of Employers,
Including Organizations,
Individuals, Attorneys, or
Other Clients Employing
Statistical Practitioners
Those employing any person to analyze conform to the expectations or desires
data are implicitly relying on the profes- of those commissioning the study or
sion’s reputation for objectivity. However, the statistical practitioner(s).
this creates an obligation on the part of the 4. Recognize that it is contrary to these
employer to understand and respect statis- Guidelines to report or follow only those
ticians’ obligation of objectivity. results that conform to expectations
Those employing statisticians are expected to: without explicitly acknowledging com-
peting findings and the basis for choices
1. Recognize that the Ethical Guidelines regarding which results to report, use,
exist, and were instituted, for the pro- and/or cite.18
tection and support of the statistician
and the consumer alike.16 5. Recognize that the inclusion of statisti-
cal practitioners as authors, or ac-
2. Recognize that valid findings result knowledgement of their contributions
from competent work in a moral to projects or publications, requires
environment. Employers, funders, their explicit permission because it
or those who commission statistical implies endorsement of the work.
analysis have an obligation to rely on
the expertise and judgment of qualified 6. Support sound statistical analysis
statisticians for any data analysis. This and expose incompetent or corrupt
obligation may be especially relevant in statistical practice.
analyses that are known or anticipated 7. Strive to protect the professional
to have tangible physical, financial, or freedom and responsibility of statis-
psychological impacts.17 tical practitioners who comply with
3. Recognize that the results of valid sta- these Guidelines.19
tistical studies cannot be guaranteed to

9 Ethical Guidelines for Statistical Practice

American Statistical Association
Discussion
1 The use of statistics in medical diagno- that are clearly defined for the stage
ses and biomedical research may affect (exploratory, intermediate, or final) of
whether individuals live or die, whether analysis before looking at those data that
their health is protected or jeopardized, will be decisive for that stage and tech-
and whether medical science advances nical criteria to justify both the practical
or becomes sidetracked. Life, death, and relevance of the study and the amount of
health may be at stake in statistical anal- data to be used.
yses of occupational, environmental, or
transportation safety. Early detection and 4 Running multiple tests on the same data
control of new or recurrent infectious set at the same stage of an analysis increas-
diseases depend on sound epidemiological es the chance of obtaining at least one in-
statistics. Mental and social health may be valid result. Selecting the one “significant”
at stake in psychological and sociological result from a multiplicity of parallel tests
applications of statistical analysis. poses a grave risk of an incorrect conclu-
sion. Failure to disclose the full extent of
Effective functioning of the economy de- tests and their results in such a case would
pends on the availability of reliable, timely, be highly misleading.
and properly interpreted economic data.
The profitability of individual firms de- 5 For example, address the multiple
pends in part on their quality control and potentially confounding factors in obser-
market research, both of which should rely vational studies and use due caution in
on statistical methods. Agricultural pro- drawing causal inferences. The fact that a
ductivity benefits greatly from statistically procedure is automated does not ensure
sound applications to research and output its correctness or appropriateness; it is also
reporting. Governmental policy decisions necessary to understand the theory, data,
regarding public health, criminal justice, and methods used in each statistical study.
social equity, education, the environment,
national defense, and security depend in 6 Preferably, authorship order in statisti-
part on sound statistics. cal publications should be by degree of
intellectual contribution to the study and
2 Misconduct is not limited to instances of material to be published, to the extent that
plagiarism and data fabrication or falsifi- such ordering can feasibly be determined.
cation. More broadly, misconduct includes When some other rule of authorship order
all professional dishonesty, by commis- is used in a statistical publication, the rule
sion or omission, and, within the realm of should be disclosed in a footnote or endnote.
professional activities and expression, all Where authorship order by contribution is
harmful disrespect for people, unautho- assumed by those making decisions about
rized use of their intellectual and physical hiring, promotion, or tenure, for example,
property, and unjustified detraction from failure to disclose an alternative rule may
their reputations. improperly damage or advance careers.

3 Typically, each study should be based on 7 This may sometimes require divestiture
a competent understanding of the sub- of the conflicting personal interest or
ject-matter issues and statistical protocols withdrawal from the professional activity.

10 Ethical Guidelines for Statistical Practice

American Statistical Association
Examples where conflict of interest may 12 Ensure adequate planning to support
be problematic include grant reviews, the practical value of the research, validity
other peer reviews, and tensions be- of expected results, ability to provide the
tween scholarship and personal or fami- protection promised, and consideration of
ly financial interests. all other ethical issues involved.

8 For the general public, convey the scope, 13 These recommendations may be
relevance, and conclusions of a study based on prospective power analysis,
without technical distractions. For the the planned precision of the study end-
professional literature, strive to answer the point(s), or other methods to ensure
questions likely to occur to your peers. appropriate scope to either frequentist or
Bayesian approaches. Study scope also
9 For example, disclose any significant fail- should take into consideration the feasibil-
ure to follow through fully on an agreed ity of obtaining research subjects and the
sampling or analytic plan and explain any value of the data elements to be collected.
resulting adverse consequences. Address
the suitability of the analytic methods and 14 Appropriate approval is not necessarily
their inherent assumptions relative to the given by the research subjects themselves.
circumstances of the specific study. Iden- Approval may be required from a parent
tify the computer routines used to imple- or legal guardian who must consent, and
ment the analytic methods. the actual participant may be required to
assent in some cases. This includes legal
1⁰ Statisticians are encouraged to partici- guardians for minors and adults who are
pate in professional activities contributing unable to give consent. In animal studies,
to the improvement of the community statisticians should enquire as to proper
and to work that elevates the statistical forms of owner or institutional consent for
profession in the United States and the animals that are part of a research project.
world. It is recognized that the ability to
do pro bono work may be limited by the In all cases, avoid or minimize the use of
conditions of the statistician’s employment deception. Where it is necessary and pro-
and personal situations, but statisticians vides significant knowledge—as in some
should be open to opportunities for pro psychological, sociological, and other
bono and other work, including service to research—ensure prior independent eth-
the local community or to international ical review of the protocol and continued
organizations. Service to the profession— monitoring of the research. Where full
including service on ASA committees, sec- disclosure of study parameters to subjects
tions, and chapters—is also encouraged. or other investigators is not advisable, as
in some randomized clinical trials, gen-
11 Statistical methods may be broadly erally inform them of the nature of the
applicable to many classes of problem or information withheld and the reason for
application. Statistical innovators may withholding it. As with deception, ensure
well be entitled to monetary or other independent ethical review of the protocol
rewards for their writings, software, or and continued monitoring of the research.
research results.

11 Ethical Guidelines for Statistical Practice

American Statistical Association
15 All research is constrained by resources, expose incompetent or corrupt practice
including the available funding, timing, before it can cause harm to research sub-
and personnel. Thus few, if any, studies jects or society at large.
can achieve the highest standards imag-
inable. However, the statistician should 18 Any measures taken to ensure a partic-
always strive to achieve the most valid ular outcome will lessen the validity of the
results with the resources available and analysis. Pressure on a statistical practi-
avoid working on studies likely to produce tioner to deviate from these guidelines is
misleading or meaningless results. likely to damage both the validity of study
results and the professional credibility of
16 This safeguard will lower your risk of the practitioner.
learning only after the fact that you have
collaborated on an unethical study. 19 Within organizations and within pro-
fessions using statistical methods gener-
17 In cases of conflict, statistical practi- ally, statistics practitioners with greater
tioners and those employing them are prestige, power, or status have a responsi-
encouraged to resolve issues of ethical bility to protect the professional freedom
practice privately. If private resolution and responsibility of more subordinate
is not possible, recognize that statistical statistical practitioners who comply with
practitioners have an ethical obligation to these guidelines.

12 Ethical Guidelines for Statistical Practice

American Statistical Association