EBP1 Week 3 Material Flashcards
Research Studies, RCTs and Confounding, [Non]Probability, Experimental Design, Critical Appraisal
What are the three types of statistical studies?
Sample Study, Observational Study, Experimental Study
What is the function of a sample study?
To estimate value of a parameter for a population by taking a random sample of it
What is the function of an observational study?
Understand how two parameters in a population relate to one another
What is the function of an experimental study?
Establish cause and effect (the basis of SM)
What two groups are created in an experimental study?
Control and Experimental
Was is the definition of bias in an experimental study?
Error in methodology that results in a systematic deviation of results away from the “truth”
What is the meaning behind Lost to Follow Up?
Missing participants and figuring out how that affects your results
What is, more or less, the most important component to an experimental study?
Random Assignment
Define a confounding variable.
an alternative explanation on an observed relationship (Shark attacks = ice cream sales)
What is the population in a study?
Persons, objects, events that meet the criteria
What is the target population?
The larger population to whom results would be generalized to
What is the accessible population?
The actual population available to be chosen for study
Differentiate between sampling bias and sampling error.
Sampling bias is systematically misrepresenting population while sampling error is randomly misrepresenting the population
What are the key differences between probability samples and nonprobability samples?
Probability samples are randomly chosen, can generalized to a population, can estimate sampling error while nonprobability is the opposite
What are the four types of probability sampling?
Simple random, Stratified random, Systematic, and Cluster
What is a systematic probability sample?
Selecting certain one (ex: every 10th person)
What is a stratified random probability sample?
Specifying a number from each category
What is a cluster probability sample?
Multilayer/Stage
What are the five types of non-probability sampling?
convenience, consecutive, volunteers, quota, purposive
What is a convenience NP sample?
Basis of availability
What is a consecutive NP sample?
Recruiting all people to meet criteria
What is a volunteer type NP sample? Or better, how does one promote one?
Flyers/ads
What is a quota NP sample?
Specifying a certain number to each category but not at random
What is a purposive type NP sample?
Subjects were hand-picked by specific criteria (case series)
Name the four types of experimental research trials discussed in this section.
Randomized Trial, Clinicial Trial, RCT, Randomized clinical trial
What is the purpose of experimental research?
To make inferences on cause and effect
What is an extraneous variable?
A factor not related to the purpose of a variable
What is another name for an extraneous variable?
Nuisance variable
What happens to an extraneous variable if it is uncontrolled?
Can become a confounding variable
What is experimental research designed to control for?
Confounding variables
What is the most effective design strategy for ruling out extraneous variables?
Use of control group
What are the three essential components of experimental research?
- Include a control/comparison group 2. IV is manipulated by the experimenter 3. subjects are randomly assigned to groups
What is the meaning behind random assignment?
Each subject having an equal chance of being assigned to any group
What is an important component of random assignment?
Concealing allocation (sealed envelopes, locked cabinets)
What does blinding control?
bias
Who are the four types of people that can be blinded?
- patient 2. treater 3. measurer 4. data analyzers
What are the two types of blinding?
Single blind and double blind
What is single blinding?
Measurers/outcome assessors are blinded
What is double blinding?
Measurers/Outcome assessors and patients are blinded
What component of an experiment must be equal across all groups?
Intervention
What are the two ways to account for missing data?
Per protocol analysis and Intention to Treat Analysis
Discuss Per Protocol Analysis.
You analyze the people who completed the study (toss out, more or less)
Discuss Intent to Treat Analysis.
You analyze the group that they were assigned to and impute/estimate missing data but you do not discard them. important to try to find out why subject went missing
Discuss the five design strategies for controlling inter-subject differences (cofounders).
- make variables homogenous (only 2 groups, divide by age, sex) 2. Build into study as IV (m/f, age blocks) 3. matching (age and sex) 4. Using subjects as their own control (within-subject design) 5. Analysis of covariance (statistically adjust DV based on differences in covariate)
What are the four types of experimental design validity?
External, Construct, Internal, Statistical Conclusion
Which two types of experimental design validity were discussed to be the most important?
Internal and External
What are the threats to external validity? [3]
Selection (sample doesn’t represent pop), setting (what setting it took place e.g. military v. civilian), History (how long ago was study completed?)
What is the meaning behind internal validity?
Is there a causal relationship between IV and DV
What is the meaning behind external validity?
Can the results be generalized to other people, settings, times?
What is the meaning behind construct validity?
What theoretic constructs can be generalized?
What is the meaning behind statistical conclusion validity?
What is the relationship between IV and DV?
What are the five types of experimental design?
True experimental, Quasi-experimental, Designs for independent groups, designs for repeated measures, RCTs, Mixed Design
Which type of design is considered to be the gold standard?
RCT
What is a true experimental study?
IV is manipulated by researcher, random assignment into at least two groups
What is a quasi-experimental study?
May lack randomization and comparison group, they accommodate for limitations of natural settings where random assignment isn’t available/ethical
What is between-subjects design?
Different people in each level of IV (Independent Groups)
What are the two types between subjects design?
Single-factor and Multi-factor
What are the two types of single-factor design?
Pretest-Posttest and Posttest only
What are the advantages to posttest only?
Less laborious and costly, appropriate when pretest is impractical and may skew results
What are the disadvantages to posttest only?
You do not know if the groups were equivalent at baseline
What are the ways to measure multi-factor designs? [3]
- Interaction effect (do scores of A depend on levels of B) 2. Randomized block (one IV is controlled) 3. Multi-factor (2 or more IVs what is the interaction between the levels of ABC)
What are the three types of within- subjects design?
- single factor 2. crossover 3. multifactoral
Describe single factor within-subjects design.
Subjects are exposed to all levels of IV (they get all treatments), no control group. Downside: outcome may depend on the order of measurement (order effects) so to combat that: randomize the order, counterbalance, or use Latin square
Describe the crossover design of repeated measures.
Special type of single factor - used when 2 levels of IV and condition is stable over time. Two groups: receive A then B and vice versa. Also affected by order effect so use washout period (sufficient period of time to eliminate effects of treatment)
Describe the multifactoral Within Subjects Design.
subjects exposed to two or more IVs
Describe a Mixed Design.
At least one IV is within-subjects and one IV is between subjects
When it comes to interventional studies, what two things do researchers often distinguish between?
Efficacy and Effectiveness
What is efficacy?
Benefit of an intervention in a highly controlled environment (vs standard)
What is effectiveness?
Benefit of an intervention to under pragmatic “real-world” conditions
What are the three steps in critical appraisal?
Are the results valid? What are the results? Will the results help care for my patients?
What are the seven questions you need to ask about an article to determine its validity?
- were they randomly assigned? 2. What allocation concealed? 3. Were baseline groups similar 4. Was Intention to Treat analysis used? 5. Was Follow up sufficiently long and complete? 6. Were patients, clinicians, study personnel blinded? 7. Was intervention equal/were all groups treated equally?
