PSYCO105 Lecture 6: Error and bias in research
18 Jan 2019
School
Department
Course
Professor
PSYCH 105: Individual and Social Behaviour (18th January 2019)- Lecture 6
➢ Error
- Error is due to random variability in participants (mood and personality) and environment.
- Error refers to random fluctuations in results that the researchers cannot control.
- The consequences of error can cancel out each other when calculating average. *
- Additionally, error can be considered through statistical calculation.
- Errors in an experiment are inevitable.
➢ Bias
- Bias is a more serious problem than error.
- It is a non-random effect and can not be accounted for by statistics
- It is caused by factors extraneous to the research hypothesis
- Bias can lead to false conclusions. *
➢ Problems caused for researchers:
- Error: If there is more error, there is more variability. So, chances of finding statistically
significant results decreases.
- Bias: Can lead to false conclusions. Researcher may conclude that hypothesis is correct, but
truly it is due to an irrelevant factor.
➢ Biased Samples
- Not representative of the population
- Random assignment is the best way to avoid bias
- If not random assignment, as in the participants choose which treatment group they want to
belong to, there will be bias
- As participants will choose their group according to their preference and motive.
- Simply using a large sample doesn’t eliminate bias in an experiment.
Literary Digest’s Poll is a presidential poll. In 1936, they mis predicted the president of the
USA, despite sampling 2 million people. Their sample was not representative of the
population, as it was a time of depression and most people were not in a good financial
position. However, their sample contained people from telephone lists, the magazine
subscribers and club memberships, who were well off. This bias let to wrong conclusions.
- Most psychological research is now conducted on university students. This is a major source
of bias, as university students are educated and mostly well-off which doesn’t represent the
whole population.
- Researchers opt for university students because it is a convenient way of collecting data and
less expensive.
➢ Avoiding measurement bias
1. Reliability: (repeatability or replicability): tendency for a measure to produce the same
results when repeated. It is associated with measurement error and not bias. Low
reliability (high error) decreases the chance of getting statistically significant results.
2. Validity: how close is a measurement to the measurement in reality. It can be a source of
error and/or bias. A procedure may be reliable but not valid.
Using the length of thumb to predict ones IQ score. It may give reliable results, as in when
repeated it gives the same results, but it is not essentially valid/ true.
Document Summary
Psych 105: individual and social behaviour (18th january 2019)- lecture 6. Error is due to random variability in participants (mood and personality) and environment. Error refers to random fluctuations in results that the researchers cannot control. The consequences of error can cancel out each other when calculating average. Additionally, error can be considered through statistical calculation. Bias is a more serious problem than error. It is a non-random effect and can not be accounted for by statistics. It is caused by factors extraneous to the research hypothesis. Error: if there is more error, there is more variability. So, chances of finding statistically significant results decreases. Researcher may conclude that hypothesis is correct, but truly it is due to an irrelevant factor. Random assignment is the best way to avoid bias. If not random assignment, as in the participants choose which treatment group they want to belong to, there will be bias.
