lecture 2 - introduction to quantitative research designs Flashcards
(28 cards)
why does the research design matter?
- Efficiency
- Its a blueprint for advance planning
- Error reduction
- Reliability and evidence design dependent
- Is the design ethical?
Allows hypotheses to be tested
what is validity?
are you measuring what you set out to measure or are there confounding variables? can we establish cause and effect?
what is internal validity?
are there problems due to manipulation or other causes (variables)?
- was your study done correctly?
- can you claim c+e?
- are your measures appropriate and valid?
have you controlled for other causes?
what does external validity measure?
generalisability to a wider population, comparability with other literature
- did lots of people drop out?
- can you make generalisations? - high dropout rare questions the generalisability
- selection bias - are you choosing based on friendship/location - selection bias, is it representative?
what does no intervention mean?
no manipulation of the independent variable, you are only observing the effects of the independent variable
under no intervention, what research design is used when there is a comparison group?
- cohort study
- case-control
under no intervention, what research design is used when there is no comparison group?
- case series
- case study
- cross-sectional
look at exposed group and what happened
what does intervention mean for a research design?
manipulating independent variable - choosing the variable and level of it
experimental designs
under intervention, what research design is used when there is random allocation?
randomised controlled trial or systematic review
under intervention, what research design is used when there is no random allocation?
before and after study
interrupted time series
explain correlational designs
- look at the impact of a variable (independent) and relationship with another variable (dependent)
- participants choose independent variable (exposure to variable)
- observing and recording only
- no intervention
- ethically neutral
- correlation not causation - effects of independent variable but no evidence that it actually caused it
what are common threats with correlational design?
- recall bias - recalling exposure to independent variable e.g. when did I get COVID
- difficult to get random samples - looking for specific people
- cannot account for all confounders
- cannot prove causation only correlation e.g. self esteem -> depression or depression -> self-esteem or unknown variable causing both
- temporal relationship not always clear
- attrition bias - drop out or systematic drop out - may pass away for example
what is the case study design?
- sometimes called case report
- usually a single case - limited generalisability
- very limited causality
- hypothesis generating for related studies
- retrospective e.g. look at someone who broke leg and what happened
what is a cross-sectional design
- data collected on series of patients/participants
- single time point - only measured once
- what is happening now - only data collected
- associations between variables only not causality
what is the structure of a cross-sectional design?
a cross-sectional design comprises the collection of data on a series of variables at a single point in time
what is a case-control study?
- outcome precedes exposure
- go back in history to try and understand how some exposure linked to disease risk
- good for rare outcomes
- recall bias?
- compares groups retrospectively to identify possible predictors of an outcome
- useful for studying rare diseases or outcomes
- often used to generate hypotheses
what are cohort studies?
- aka longitudinal study
- follow a population over time
- uses questionnaires, interviews etc
- temporal relationship between variables
- exposure precedes outcome
- won’t measure every outcome or exposure
- sample is surveyed more than once
○ Randomly selected panel
○ Cohort study has people with similar characteristics
Sample could change over time
what does a cohort study design comprise?
collection of data on a series of variables at multiple time points
what is an experimental design?
- widely used
- manipulates one variable called the independent variable to see what effect this has upon another variable called the dependent variable (dependent)
- is random assignment used
- yes - randomised or true experiment
- no - is there a control group or multiple measure - yes - quasi-experiment
no - non-experiment
what are quasi-experimental approaches?
- before and after study
- interrupted time series
what is the before and after study?
single group measure before, then intervention, then measured again
- learning effects - might do better second time if they learn task
- easier with smaller samples
- no control group
- cheap and easy
- not measuring any confounding variables
what is the interrupted time series?
- series of measurements interrupted by intervention
- single group measured before, then intervention, then measured again
- but with multiple measurements before and after
- can examine seasonal trends
- useful for examining policy impact
- more robust to bias than before-after studies
- cheap if have access to routine data
what are randomised controlled trials
- group performs task (no intervention), perform it again later
- participants randomly allocated intervention or control
- prospective
- gold standard effectiveness
- can determine cause and effect - controlling other variables
BUT - expensive - lot more participants and 2 groups
- not always ethical - can involve medication - unethical for control group to not take medication
what is the criteria for causality?
- theoretical plausibility
- co-variation
- time-order
- non-spuriousness
