Lecture 2 Flashcards
Why is a good experimental design important?
As behaviour and cognition follow certain predictable patterns, or be able to infer associations and causality we require a good experimental design
Why is figuring out psychological principles of human behaviour and cognition so difficult?
Some reasons are exceptions, variance, researchers are human and pattern seekers, observing affects the observed, behaviour and cognition are multiply determined
Noise
Noise - extraneous variables in an experiment that influences the dependant variable but that is even,y distributed across the experimental conditions.
Noise doesn’t threaten validity but it decreases the ability to detect an effect statistically
How do you work out effect?
Effect = signal+noise divided by noise
How do you minimise noise?
Good research design
What is reliability?
Is the measure consistent/stable?
Actual measure = hypothetical true score + measurement error
What is validity?
Does the measure or study truly asses the variable interest?
Internally valid vs externally valid
What are confounds?
Nuisance variables that vary systematically with the IV and influence the DV in consistent ways
Threatens internal validity
What are the three types of confounds?
Person
Operational
Procedural
What is a person confound?
When individual differences covariant with the independent variable
Threatens internal validity
Can randomly assign individuals, can try and use matching e.g. ensure same amount of males and females in each group
What are operational confounds?
When a measure designed to asses a particular construct inadvertently measures something else as well
Threatens construct validity
One option is to refine the operational definition
What are procedural confounds?
The researcher manipulates one thing (IV) but inadvertently another thing covaries with it
Threatens internal validity
One option is to repeat study while controlling this variable
What are within subject confounds?
Carryover effects:
Practice, fatigue, order, priming, interference, framing
What threats are there to internal validity in studies extending over time?
History - other things may have happened in the world
Maturation - natural changes to the subject, not just age, e.g. symptom fluctuation
Instrumentation - the measuring instrument has changed or is not identical
Attrition (subject mortality)
Regression to mean - tendency for people who receive very hi or low scores to score closer to the mean in subsequent measure - researchers consistently underestimate this (measurement = true score + noise)
How do you minimise confounds?
Random assignment,
control group: random assignment matching, undergo the same except when using placebo
Placebo effect sham operation
What are common approaches to minimising confounds?
Double blind placebo controlled (as opposed to open label)
Randomised controlled trials
Counterbalancing - complete (n!) or partial - Latin square, random, reverse (abccba)
Control conditions
Pre test measurements
Pilot studies
What is artefact?
When a variable is held constant in a study but influences the relation between the independent and dependent variables.
This is a threat to external validity. It can be reduced through random selection and maximising experimental realism.
What is experimenter bias?
Inadvertently doing something leading subjects to behave in ways that confirm (or disconfirm) the hypothesis e.g treating participants differently or selective observations
Reduce it using double blind procedures and strict protocol adherence
What are subject biases?
Hawthorne effect
Participant reaction
Evaluation apprehension
Good subject effects
How do you reduce subject biases?
Single blind procedures
Manipulation checks e.g. mood ratings following mood induction
Reducing demand characteristics - disguise the hypothesis from subjects
Structured debriefing e.g. so what do you think the study was looking for?
Providing a cover story
Guarantee anonymity
Use of unobtrusive or indirect measurements
What is selection/subject selection bias?
The relationships between certain subjects and environmental variables that may arise when people with different biological, behavioural and psychological characteristics select different types of environments/situations
What are sampling biases?
When the sample does not correctly represent the research population (often critical for survey research)
What are ceiling and floor effects?
Insufficient variation in performance because the task is too easy or difficult
E.g. everyone circles 1 on likert scale or scores 95% and above on a test
Golden rules for the aspiring researcher
Design best study you can
Pay attention to detail
Document everything
Be able to justify everything
