Research Methods Flashcards
Why is experience not a good basis for ideas?
We only see one version of a situation rather than every possible combination.
What are two reasons why intuition can be flawed?
- Overconfidence
- Confirmation bias
Theory-data cycle
Develop theories of behaviour, which lead to hypotheses, which are tested by collecting data.
Data consistent with a hypothesis increase support for a theory; data inconsistent with a hypothesis decrease support for a theory.
Variable
Something of interest that can take on different values from person to person.
Measured variable
Researcher observes and records value in numeric form.
Manipulated variable
Researcher assigns each individual to one level of a variable or another.
Operationalizing a variable
Turning a variable into a specific number or value.
What are 3 methods for operationalizing measured variables?
- Self report
- Physiological
- Observation
What are the 3 kinds of research?
Surveys
Descriptive
Correlational
What is a survey? What do they try and find out? How do they do it? How effective are they?
a. What do people do on average?
b. Descriptive/self-report
c. Sample
d. If random sampling is used, a sample as small as 1500 people can generalize to the population of interest
What are the two types of descriptive study?
a. Naturalistic observation: Use observational measures to record behaviours unobtrusively
b. Case studies: Study one or a few people, often extensively and over a long period of time
What is a correlational study?
a. Measure two or more variables and quantify relationships between them
b. Scatterplots: Assign measured variable to each axis and plot one dot to represent where each individual falls. If correlation positive, high values on one variable go with high values on the other. If negative, high values on one go with low values on the other. If correlation is zero, values are not systematically related at all.
What can and can’t correlation let us predict?
• Positive or negative correlation allows us to predict people’s standing in one variable from the other
• Correlation necessary but not sufficient for causation
What are independent and dependent variables in an experiment?
• One variable (the hypothesized cause) is manipulated, often by randomly assigning individuals to two groups. The other variable (the hypothesized effect) is measured for all individuals.
• Independent variable: the manipulated variable
• Dependent variable: the measured variable
What 3 things must an experiment do to establish causation?
○ Covariance: independent variable groups differ on the dependent variable
○Temporal precedence: Prove independent variable occurs first, followed by the dependent variable. Because experiments manipulate independent variable and then measure the dependent, they almost always meet this criterion.
○ Rule out alternative explanations. Random assignment usually rules out lots. But must carefully control any other extraneous variables too.
Random assignment vs random sampling
• Random assignment is used only in experiments–random method like coin flip used to assign participants to one experimental condition or the other. Helps rule out alternative explanations.
• Random sampling can be used in any type of study, but most commonly used in surveys. Lets us generalize results from sample to population of interest.
What are the advantages and disadvantages of descriptive, correlational, and experimental research?
- Descriptive: tells us what people typically do, but not when they do it or who is most likely to do it
- Correlational: how variables are related to each other, variables that can’t be manipulated
- Experimental: support causal claims
What is validity?
The appropriateness of the claim or conclusion
How do you assess construct validity?
- Validity: the appropriateness of the claim or conclusion
- Construct validity: how well the study operationalized its measured and manipulated variables
- Operationalizations of measured variables should be reliable (consistent) and should measure what it’s supposed to measure
How do you assess external validity?
- External validity concerns who was included in the study and how they were selected. If a study is externally valid then its results can be generalized from the sample to the population of interest.
- Random samples have external validity in as much as they can be generalized to the population they were sampled from
- Studies conducted on one type of person may not be externally valid if they were meant to understand different types of people
When does a study have internal validity?
- Internal validity is mainly relevant when a study is trying to support a causal claim. A study is internally valid if the researchers have ruled out any relevant alternative explanations for the relationship.
- Well conducted experiments tend to have good internal validity, because of random assignment and carefully controlled independent variables
- Correlational studies don’t tend to have good internal validity because they can’t rule out alternative explanations for the association between the variables
How can we evaluate claims in popular media? (5)
- What am I being asked to believe?
- What evidence is there?
- How strong is the result?
- Are they making a causal claim?
- Has the study been replicated?
Frequency distribution
Bar graph that shows the possible values of a variable and how many people are at each value
Central tendency
Average, described by mean, median, and mode
Variability
We describe how variable a batch of scores are by using standard deviation or variance
Effect size
Quantification of the strength of the relationship between two variables
R
Correlation coefficient
Used in correlational studies graphed with a scatterplot
R=.1 small
R=.5 strong
D
Difference between two conditions.
Difference between the two group means, divided by the standard deviation.
.2 small
.8 strong
Inferential statistics
Uses sample results to estimate what’s true about the broader population
Statistical significance
Assesses the probability that an effect size found in a sample represents your population.
When effect sizes are large they’re more likely to be called statistically significant.
Meta-analysis
Combines the effect size estimates from a large number of studies on the same question to get a better estimate of the effect size of a population.
Three principles of ethical research on humans
- Autonomy: informed consent.
- Beneficence: No harm, benefit society.
- Justice: Types of people studied must be those who benefit from it.
What environmental factors for animal research must researchers follow guidelines about?
Living space, food, housing conditions, stimulation.
What are the three Rs of ethical research on animals?
- Replacement: find alternatives to animals wherever possible
- Refinement: adjust procedures to minimize distress
- Reduction: use designs that require fewest animals possible
Animal welfare vs animal rights
Animal welfare: Animals should be protected from suffering
Animal rights: Animals shouldn’t be used bc they can’t give informed consent
Why is replication important?
It helps establish whether a study’s original findings reflect a true effect. Some notable replication efforts work with large and diverse samples.
What are the two reasons a study may fail to replicate?
- Not a true effect
- Problems with the replication study
When are studies less likely to replicate?
- Questionable research practices
- HARKing
- P-hacking
- Underreporting nonsignificant results
What two cultural factors strengthen reliability and validity?
- Open science
- Preregistration of hypotheses
