Statistics turns data into useful insight, but data often represents real people with rights, risks, and expectations. Data ethics is the practice of collecting, analyzing, sharing, and storing data in ways that are fair, honest, and respectful. Privacy matters because even harmless-looking records can reveal sensitive information when combined with other data.
Responsible statistics means producing knowledge while protecting the people behind the numbers.
Ethical data handling begins before analysis, with clear consent, limited collection, and a specific purpose. During analysis, statisticians reduce risk using methods such as anonymization, aggregation, secure storage, and bias checks. Results should be reported honestly, including uncertainty, limitations, and possible harms.
Good data work balances insight with protection, so decisions based on statistics are both useful and trustworthy.
Key Facts
- Responsible Statistics = Insight + Protection
- Collect only the data needed for a clear purpose, often called data minimization.
- Informed consent means people understand what data is collected, how it will be used, and what risks exist.
- Anonymized data removes direct identifiers, but re-identification can still occur when datasets are linked.
- A sample proportion is p-hat = x/n, but ethical reporting also requires context, uncertainty, and limitations.
- Bias can enter through sampling, measurement, missing data, algorithms, or interpretation.
Vocabulary
- Data ethics
- Data ethics is the study and practice of using data in ways that are fair, transparent, lawful, and respectful of people.
- Informed consent
- Informed consent means a person knowingly agrees to data collection after being told the purpose, uses, risks, and choices.
- Privacy
- Privacy is a person's ability to control access to information about themselves.
- Anonymization
- Anonymization is the process of removing or changing identifying details so data is less likely to be linked to a specific person.
- Bias
- Bias is a systematic error that makes data, analysis, or conclusions unfairly favor one group or outcome.
Common Mistakes to Avoid
- Assuming removing names makes data completely anonymous. This is wrong because age, location, dates, and other details can be combined to re-identify people.
- Collecting extra data just in case it might be useful later. This is wrong because unnecessary data increases privacy risk and may violate the purpose people agreed to.
- Reporting a statistic without explaining how the data was collected. This is wrong because sampling methods, missing data, and measurement choices can strongly affect the result.
- Treating an algorithm as neutral because it uses numbers. This is wrong because algorithms can reproduce bias from training data, design choices, or unequal measurement.
Practice Questions
- 1 A school survey has 500 responses. Before sharing the dataset, the researcher removes names from all records, but 35 records still include exact birth date, ZIP code, and club membership. What percent of the records may still have high re-identification risk?
- 2 A health app collected 12 variables from each user, but only 7 are needed to answer the research question. If the dataset contains 20,000 users, how many unnecessary data values were collected in total?
- 3 A city wants to publish a map of disease cases by neighborhood. Explain why showing individual home locations would be ethically risky, and describe one safer way to share useful information.