1b. Quantitative research methods Flashcards
What is the goal of an experiment?
To determine a cause and effect relationship between two variables
Confounding/extraneous variables
They contribute to bias as they could potentially influence relationship.
How do you control confounding variables?
They are eliminated or kept constant
Lab/True experiment
When independent variable is manipulated to study effect on dependent variable
Characteristics of lab experiments
- High degree of control/standardized
- Cause-and-effect is established
- Random allocation to remove demand characteristics
- Quantitative data is collected
- May not be realistic if there is too many confounding variables
Field experiment
Studies that occur outside lab conditions
Characteristics of field experiments
- Cannot control confounding variables
- Cannot be easily replicated
- High ecological validity
Quasi experiment
IV is not manipulated. They are naturally pre-existing differences like age, gender, ethnicity
Characteristics of quasi experiments
- No cause and effect established
- Participants are not randomly allocated
- Setting can be lab or field
- Implies casual relationship
Natural experiment
Behavior is measured before and after a variable is introduced eg. behavior of smokers after cigarettes were banned
Characteristics of natural experiments
- No cause and effect established
- Participants are not randomly allocated
- Setting can be lab or field
- Implies casual relationship
Hypothesis
Prediction of how IV will affect DV
Null hypothesis
States there is no significant difference i.e. no relationship between IV and DV
Experimental hypothesis + types
Predicts a relationship between IV and DV
- One tailed: Direction of relationship is specified
- Two tailed: No direction specified
Difference between aim and hypothesis
- Aim is just IV, DV and target population
- Hypotheses includes operationalized IV and DV
What are limitations of experiments
- Artificial set-up
- May not reflect real life
- May lack ecological validity
What are the 3 sources of bias or error?
- Participant
- Researcher
- Sampling
Demand characteristics
Occur when participants act differently because they know they’re in a study
Expectancy effect
Participants attempt to guess researcher’s hypothesis with the aim of helping the researcher. (They might act a certain way or try to give right answers)
Screw you effect
Participants attempt to guess the researcher’s hypothesis but only in order to destroy credibility of study
Social desirability effect
Participants answer in a way that makes them look good to the researchers. Done to avoid embarrassment or judgement
How can you avoid demand characteristics?
Single blind studies. Experiment where researchers know which participants are receiving which treatment but participants don’t know which condition they are in
Participant variability and how to control
Characteristics of sample affect DV and can only be controlled by randomly allocating people to groups
When might sampling bias occur and what will it result in?
- When psychologists use non-probability sampling technique (everything except random sampling), there may be bias
- May cause some members to be less likely to be included than others
- Participant variables may not be representative and hence influence outcome
Researcher bias
When experimenters see what they are looking for
Types of researcher bias
- Confirmation bias: Paying attention to info that agrees and discount information that contradicts it
- Publication bias: Researchers may manipulate data for quicker publications as it is expected of them to publish for credit
How to control for researcher bias
Double blind control. Both participants and researcher do not know who is receiving which treatment
Research designs
Overall strategy that a researcher uses to investigate the research problem
Research designs: Repeated measures
One sample of participants receive all conditions of an experiment
Research designs: Strengths of repeated measures
- Individual is only compared to themselves so participant variable is controlled
- Requires lesser number of participants
Research designs: Limitations of repeated measures
- Order effects: Differences in participant responses result from order in which they participate in conditions
- Demand characteristics are not controlled
Research designs: 3 types of order effects
- Fatigue effect: Decrease in performance in later conditions is because they are tired or bored
- Interference effect: When first condition may influence the outcome of the second
- Practice effect: Improvement in performance as they have developed the skill
Counterbalancing
Technique used to deal with order effects. eg. Sample is divided in half and both groups attempt both conditions but in different orders
Research designs: Independent measures
Random allocation of participants
Research designs: Pros and cons of independent measure
Pros: - Less chance of order effects - Less chance of demand characteristics as participants cannot guess (only exposed to one condition) Cons: - Affected by participant variables - Requires a lot more participants
Research designs: Matched pair design
Matches participants who have equal characteristics and then divides according. Usually depends on DV (eg. level of memory)
What is not matched pair design?
Matching to ensure there are no extraneous variables like age, gender and race
When is matched pair design used?
When there is small groups as random allocation will not be sufficient for group equivalence
Cross sectional study
Researcher compares two or more groups on a specific variable at a specific time i.e. short-term
Longitudinal study
Where a researcher measures change in an individual over time
Retrospective study
Participant is asked about past behaviour
Limitations of retrospective study
- Relying on someone’s memory
- Not possible to verify information
- Social desirability effect
- No cause and effect
Prospective study
Measures a variable at beginning and then watches effect overtime
Limitations of prospective study
- Takes much longer to carry out
- Participants may get tired, bored or drop out
Reliability
Consistency of a measure. Degree to which a study is free of random error.
Test-retest reliability
Obtaining the same results across time with the same population
Validity
Degree to which a test measures what is claims to measure
Internal validity
When experiment is controlled to ensure only IV is affecting DV. Higher degree of control= higher internal validity
Where can high internal validity be observed?
Lab experiments
External validity and the types
Extent to which results can be generalized beyond sample tested
- Population validity
- Ecological validity
Population validity
Describes how well sample can be generalized to target population. High when sample is
representative
Ecological validity
Looks at experimental environment and determines how much it influences behaviour. Are they representative of conditions in the wider world?
Where can high ecological validity be observed?
In field experiments
Construct validity
Characterizes quality of operationalization.
Name some threats to internal validity
- Selection: Groups are not equivalent
- History: Outside events can affect behaviour especially in long studies
- Demand characteristics
- Instrumentation: eg. Natural human errors like when observing
- Testing effect
- Experimental mortality
Correlation
Measure of linear relationship between 2 variables
Positive correlation
Increase or decrease in BOTH variables at the same time
Negative correlation
One increases while the other decreases (inverse tendency)
What is the term for no relationship between variables?
Zero correlation
Correlation coefficient
Number used to denote correlation between two variables. Ranges from -1 to 1 where the extremes are perfect correlations
Interpretation if correlation coefficient is between 0.30-0.49
Probably medium relationship
Interpretation if correlation coefficient is between <0.1
Negligible
Curvilinear relationships
When 2 variables increase or decrease but only up to a certain point. After that, there will be change in the relationship
