Methods Flashcards
What is methods?
- How a study is done
- tools → data collection
- Explanation (who, how, what, when, where, why)
- detailed & procedural
Types of studies
Experiments and correlation studies
Experiments
- testing a hypothesis
- Variables
- Manipulation of Ind
- control group
- Randomization
Manipulation
Can start to apply causality
Control group t randomization
Experiments don’t need but makes more certain of causality
Control group
- keep the same/usual care
- group to compare experimental group
- helps strengthen causality of Ind →dep
- Difficult to control humans - always error
Randomization
- no pattern
- prevents bias
- group = group (don’t want groups too different from eachother)
Types of randomization
Fixed:
- Simple
- Block
- stratified
Adaptive:
- outcome adaptive
- minimization
Simple type
Patient allocation is based on a single sequence of random assignments
Stratified type
Patients are grouped according to prognostic variables, then randomized in separate strata
Block type
Patients are first allocated to blocks, then randomized w/in blocks
Outcome Adaptive type
Patient allocation probabilities charge throughout the trial according to incoming results/health outcome data
Minimization type
Allocation depends on previous allocations, in a way that as to minimize covariate imbalance
Advantages of simple type
Simple to use, unpredictable
Disadvantages of simple type
Not equal size trials at all times, can’t assure balance on outcome predictors, especially in small trials.
Advantages of block type
Equal size trial at all this
Disadvantages of block type
Assignment may be predicted if small size blocks that don’t vary are used, especially if binding can’t be implemented. Can’t assure balance on outcome predictors, especially in small trials
Advantages of stratified type
Equal site trial if combined W / blocking. Assure balance on outcome predictors, especially in small trials
Disadvantages of stratified type
Danger for overstratification o imbalances due to incomplete blocks. Prediction of allocation when small site blocks in unblind trial is possible
Drawing research design
R- random
NR- non random
O-observation
X-treatment
Correlation studies
- Examine relationship between variables
- meaningful associations
Correlation
Not = causation
Branford hill criteria
Criteria that has to be met to assume causation
-Strength
- Consistency
- specificity
- temporality
- biological gradient
- plausibility
- Coherence
- experiment
- analogy
T-test
If there is a lot of overlap in variance, they are not significantly different
Pvalue > 0.5
No diff between data sets
Pvalue < 0.5
Diff between datasets
Sampling
- Represent population
-Demographic ranges (inclusion and exclusion criteria)
Size
Population: N
Sample: n
Mean
Population: u
Sample: x bar
Standard deviation
Population: o
Sample: s
Look for assessment that
- reasonable Q count
- not complicated to use
- reliability and validity
- evidence based
- normative valves to put people in categories
Qualitative analysis
Descriptive, open-ended, and narrative data
Reliability and validity
Consistency and accuracy
How to get qualitative analysis
Surveys, interview, focus groups, existing documents
Steps of qualitative research
- Research Q
- Researcher’s role
- data collection
- Analysis
- Interpret
- Validity and reliability
- Report
Researchers role
Directly involved and biases
Data collection
Needs to be private and protected
Ex: video, audio, text
Saturation
Continue to do interview until not getting New info
Analysis
Descriptive (frequencies) and themes (patterns)
Logintudinal studies
Repeated observations of same variables over time
Cohort studies
Follows group of people over time to see is an expose is linked to health outcome
Cohort They share common characteristics
Cohort compared to similar group that was not exposed to factor
Several years
Cross sectional studies
Collect data from many different individuals at a single point in time
Retrospective studies
Analyzes past data to answer questions about future
Analyzes why something happened
Info is collected about past
Prospective studies
Follows people over time to see how factors affect outcomes
Follows people from present moment to see how future events will develop
Data collection as events happen
Case control studies
Compares people w/ condition to people w/out condition
