Practice, Ethics & Integrity
1. Documentation
-
Critical for 2 things:
-
Objective replication
-
Hypothesis
-
Design
-
Prediction
-
Procedures
-
Instruments
-
Contacting Participants
-
-
Checking and verification of results and conclusions
- Data
- Data manipulation
- Statistics
- Interpretation
-
-
Available On line
Instruments & materials
the instruments used to measure or manipulate the variables of interest, and the procedures or instruments used to control or measure extraneous variables.
-
written or verbal instructions
-
text used in email communications
-
consent forms
-
documents containing debriefing information
-
Instruments
- how many questions a scale in a questionnaire
- what the response options are
- what the range of possible scale scores is
- an example item is often provided
-
research protocol 实验流程方案
the order and manner in which materials or information were presented to the participants and the way procedures were implemented.
- the moment of first contact
- how a participant is recruited with what information and by whom.
- how consent is obtained
- what instruction or information is given, in what order
- what help is provided if participants don’t understand instructions or they behave in an unexpected manner.
- how participants are debriefed about the research after their participation.
-
preregistered 备案
-
pilot study
-
2. Data Management
-
Store
-
code book: information or metadata about what the variables mean.
- what property each variable measures
- what values mean
- what the range of possible values is
- missing value: what values are used to denote if participant did not supply relevant information
-
original data
- paper questionnaires filled out by participants
- video material used for observational coding
-
duplicate files
-
-
Process
-
Recording
-
Computing sum scores
-
-
Analyze
-
Syntax file 语法文件
a simply programming file that can be used to reproduce all the computations and statistical analyses at the push of a button.
- checking and replicating results
-
3. Unethical studies
- Tuskegee syphilis study 塔斯基吉梅毒研究
- Guatemala study 危地马拉研究
4. Ethics towards participants
-
Institutional Review Boards (IRBs) 伦理审查委员会
-
Three Principles:
-
Respect 尊重
-
Respect for the participant’s autonomy.
- Voluntary consent vs. Coercion
- extremely large financial reward
- special treatment to diseases
- students for course credit
- Voluntary consent vs. Coercion
-
Consent form
-
Often, some form of deception is necessary to control for reactivity and demand characteristics.
-
Omission: the goal of the study is not stated or formulated in very general vague terms.
-
Active deception: a cover story.
-
False feedback
-
!!danger!!: perseverance effect 固化效应
participants are still affected by the deception even after they’re debriefed and the deception is revealed and explained to them.
-
-
-
-
Beneficence 善行
- the participants should not be harmed.
- cost benefit analysis
- individual level
- broader level
- invasion of participant’s privacy
- solution: anonymity of data should only be promised if identifying information is deleted.
- who will have access?
- what it will be used for?
- cost benefit analysis
- the participants should not be harmed.
-
Justice 公正
- cost and benefits of research should be divided reasonably, fairly and equally over potential participants.
-
5. Research Integrity
- Openness, transparency and critical and systematic empirical testing.
-
Fraud
-
data were either fabricated or falsified to provide false support for the researcher’s hypothesis.
-
Fabrication 伪造
- Data were never collected, but were made up.
-
Falsification 篡改
- existing data were illegitimately altered
-
-
-
Plagiarism
- a substantial scientific contribution is presented as one’s own by copying original text, concepts or data of others without referring to the original source.
- new type of plagiarism—self-plagiarism 自我抄袭
- new type of plagiarism—self-plagiarism 自我抄袭
- a substantial scientific contribution is presented as one’s own by copying original text, concepts or data of others without referring to the original source.
-
Conflicts of Interest
- researches are funded by parties of interest.
-
Personal Values
- strong convictions or personal values can blind researchers to their data and valid critiques.
6. Questionable Research Practices
-
Practices that are acceptable if they’re implemented objectively and responsibly. But they can be abused to obtain more favorable results.
-
Harking 事后假设
- hypothesizing after the results are known
- hypothesis is adapted to fit the observed data.
- a priori hypothesis can be considered relatively strong support for a hypothesis
- Prediction is much harder.
- a posteriori hypothesis formed after the fact.
- it’s easy to find an explanation that fits a specific result.
- the original hypothesis is not confirmed.
- hypothesizing after the results are known
-
P-Hacking P值操纵
- data manipulation or selection that makes the results(p-value) more favorable.
- The golden rule of p-hacking is that as long as data selection and manipulation are reported and arguments are provided, the reader can judge for himself whether these choices are justified.
-
Cherry-picking 摘樱桃
- reporting only results that are favorable and significant.
-
Selective Omission 选择性遗漏
- the omission of non-significant results and results that contradict the hypothesis.
-
Data snooping 数据探测
- the collection of data exactly until results show a favorable p-value.
- sample size should be determined beforehand based on non-arbitrary estimates of the expected effect size of a treatment and the required confidence level
-
-
as long as choices are reported, they should be discussed and their influence on results can be evaluated.
7. Peer Review Process
8 Dissemination Problems
-
Side effect: publication bias
- favoring the publication of new and exciting significant results.
-
file-drawer problem
- it’s more rewarding to do a new study in hope for positive results and leave the null results study in a file drawer.
-
solutions:
- preregistration of the research question, the design and statistical plan.
- eliminate harking, cherry picking and selective omission.
- basing the decision to accept or reject a manuscript on preregistered research proposals instead of completed studies.
- eliminate publication bias and the file drawer problem.
- the data and statistical analysis are made publicly available.
- eliminate p-harking, fraud
- preregistration of the research question, the design and statistical plan.