W8 Best Research Practices

Framework: Australian Code of Responsible Conduct of Research; supports high-quality research, credibility, and trust in research.
Responsibilities: universities and researchers
- Researchers: conduct honestly and ethically, respect rights of those affected, promote responsible practice, disseminate findings responsibly.

National Statement on Ethical Conduct in Human Research (updated 2018): ethical conduct = doing the right thing and the right spirit (abiding respect for others).
UniSA requirement: obtain ethical approval from the UniSA Ethics Committee before data collection.

Don’t be obsessed with the p-value; avoid p-hacking and harking (hypothesizing after results are known).
Report all results, including null results.
Triple-check: verify variables, estimates, numbers, and document versions; practice good data management.
Be open and transparent: preregistration, raw data availability, and sharing analysis code.
Documentation hygiene: simple, date-based file naming to ensure you report the correct results.

The p-value is one piece of evidence, not the sole criterion.
Avoid arbitrary cutoffs like p<0.05; consider effect size and statistical power.
Non-significant results can have meaningful effect sizes; significant results don’t guarantee practical importance.
A Nature paper found about half of studied articles misinterpreted non-significance; emphasis on reporting effect sizes and context.
Always contextualize results with sample, methods, biases, and study design.

P-hacking: misuse of data to achieve a significant result.
HARKing: Hypothesizing After the Results are Known; should be avoided; preregistration helps prevent it.
A PLOS Biology study showed excess of p<0.05 values beyond what would be expected, indicating p-hacking/harking activity in some literature.

Null results should be reported to prevent publication bias and to inform replication efforts.
Use comprehensive reporting guidelines; Equator Network provides frameworks for both significant and non-significant findings.

Replication crisis: many seminal psychology findings do not replicate across labs and samples; effect sizes often shrink.
Open science: transparency about methods, preregistration, raw data, and code to improve reproducibility.

Preregistration: declare study design, measures, and statistical approach before data collection.
Share raw data and analysis code so others can verify results.
Pre-registered protocols tend to show less p-hacking; transparency correlates with more credible findings.

Avoid causal language unless the evidence supports it; correlation ≠ causation.
Consider third variables and alternative explanations for observed associations.
Example of spurious correlation: drowning rates and Nicolas Cage film releases illustrate that correlation does not imply causation.

Plan ethically and preregister; report all results; share data/code.
Be mindful of p-values, effect sizes, and power; don’t rely on arbitrary cutoffs.
Prioritize replication, transparency, and open science to strengthen trust and reproducibility.