HARking and P- Hacking
Introduction to Data Ethics
- Key Themes: Discusses the ethics of data research with a focus on two unethical practices: HARKing and p-Hacking.
- Purpose: To highlight how these practices can mislead scientific reporting and the implications for research integrity.
HARKing (Hypothesizing After the Results are Known)
- Definition: HARKing is the practice of changing or developing hypotheses after the results have been obtained. It presents an illusion that the researcher predicted the finding originally.
- Origin of the Term: Introduced in the Chapter 1 Data Ethics feature (Kerr, 1998).
- Consequences of HARKing:
- It blurs the lines between confirmatory research (testing specific hypotheses) and exploratory research (gathering data without specific expected outcomes).
- Can lead to misleading conclusions where researchers may retroactively justify a hypothesis based on results, thus misrepresenting the original investigative intent.
- Example:
- The xkcd comic strip example titled “Significant” illustrates HARKing. In this comic, scientists start with the hypothesis that “jelly beans cause acne” but only find a significant result for green jelly beans. They then create a dramatic report focusing solely on this result.
- Punchline of Comic: Highlights potential media misinterpretation with a fictitious newspaper headline “GREEN JELLY BEANS LINKED TO ACNE!”
- Ethical Scientific Reporting: Ethical researchers should report original hypotheses and all related findings regardless of their statistical significance to maintain transparency.
Publication Bias in Science
- Publication Likelihood: Research indicates that journals favor publishing studies that reject the null hypothesis over those that do not (Andrews & Kasy, 2017).
- Hypothesis Confirmation:
- Academic journals are inclined towards studies confirming preconceived hypotheses, increasing the pressure to engage in practices like HARKing.
p-Hacking
- Definition: p-Hacking combines the symbol 'p' (referring to the alpha level, typically 0.05) with the term 'hacking,' indicating manipulative practices to achieve significant results.
- Characteristics of p-Hacking:
- Researchers may decide criteria for removing outliers or other data manipulations only after looking at initial analyses.
- They might analyze data incrementally, stopping only when significant results are found instead of sticking to predetermined guidelines.
- Collecting multiple outcome variables (e.g., not just acne but also other health indicators like indigestion) but only reporting those that are statistically significant.
- Utilization of various predictor variables (e.g., different colors of jelly beans) while only reporting the predictors that yield significant findings.
- Implications of p-Hacking:
- Creates a false sense of significance in research findings and systematically undermines the reliability of reported results.
- The need for preregistration of studies to counteract these unethical practices is emphasized.
Ethical Recommendations for Reporting Results
- Clear distinction must be made between original hypotheses and others that may develop post-results.
- Report all variables and conditions of the experiment transparently to avoid misleading interpretations of surprising findings.
- Recommendations supported by literature on research practices (Hollenbeck & Wright, 2017; Simmons et al., 2011).
Conclusion
Both HARKing and p-Hacking illustrate critical challenges in maintaining ethical standards in scientific research.
There is a growing acknowledgment within the scientific community for the need for transparency and accountability in research practices to restore public trust in scientific findings.
Interactive Component: Engage in the practice of p-Hacking through a simulation study available on FiveThirtyEight for an experiential understanding of the topic.