Overview
This lecture discusses common types of bias in statistical data collection, their impact on results, and how to distinguish between them.
Non-Response Bias
- Non-response bias occurs when selected participants refuse to provide data, making it harder to get accurate information.
- This bias is increasing, making data collection (even for the U.S. Census) challenging.
- In random sampling, many people refuse to participate, so only a small fraction of those contacted give data.
- Non-response bias affects the accuracy of statistical calculations and population understanding.
Voluntary Response Samples
- Voluntary response samples occur when people self-select to participate in a survey.
- These samples are not chosen by researchers but by the participants themselves.
- People who respond often have strong opinions, leading to unrepresentative data.
- Voluntary response samples are different from non-response bias because participants choose themselves instead of being chosen by a researcher.
Deliberate Bias
- Deliberate bias involves intentionally falsifying reports or deleting unfavorable data.
- This can include deleting negative feedback or fabricating data.
- Deliberate bias makes the data unreflective of the actual population's opinions.
- Conflict of interest may influence results if the group collecting or analyzing the data benefits from certain outcomes.
- It's best to use independent statistical firms to minimize deliberate bias and conflicts of interest.
Importance of Data Collection Methods
- How data is collected (census or random sampling) matters, but these methods do not guarantee unbiased results.
- Bias can still enter data through non-response, voluntary response, or deliberate tampering.
- Always consider who conducted and funded a study to assess possible bias.
Key Terms & Definitions
- Non-Response Bias — Bias from selected participants refusing to answer or participate.
- Voluntary Response Sample — Bias from people self-selecting to participate in a survey.
- Deliberate Bias — Intentional manipulation, falsification, or deletion of data.
- Conflict of Interest — When the person or company conducting the study could benefit from certain outcomes.
Action Items / Next Steps
- Review the differences between non-response bias and voluntary response samples.
- Consider who conducted and paid for any study when assessing bias.
- Prepare for discussion on additional sources of bias in upcoming classes.