Experimental Designs Flashcards
What is non-experimental research?
- researcher observes the phenomena as they occur naturally; doesn’t intervene
- descriptive research
- correlational research
What is experimental research?
- researcher plays an active role by manipulating the IV
- examines cause and effect
- randomised controlled trial
What are the 2 types of design?
Experimental and correlational
- both designs start with an experimental hypothesis that predicts a relationship between 2 variables but the aims and methods of each approach are different
What are the different levels of the hierarchy of evidence?
- Systematic review
- Randomised control trail
- Cohort studies
- Case reports
- Expert opinion
What are the different significance of evidence levels?
- Randomised controlled studies
- Controlled longitudinal studies
- Uncontrolled longitudinal studies
- Cross-sectional studies and case studies
- Expert opinions
What are the different elements of study design?
- objectivity
- hypothesis testing
- experimental and study designs
- large representative samples
- data collection by structured instruments
- numerical data
- analysis by statistical tests
- conclusion bass on statements of probability
- validity and reliability
- generalisability
What are cohort studies?
- looking for differences between groups and try to identify variables
- looking for a causal relationship
What is required to infer causality?
- the cause must precede the effect
- the causal variable and effect variable must be associated with each other
- the relationship must not be explicable by any other cause
What do experimental designs include?
- manipulation
- control
- randomisation
What do quasi-experimental designs include?
- similar to experimental designs
- less control (no comparison group, not randomised)
- less certain that results are due to the intervention
What is manipulation?
- process by which the researcher manipulated the IV in order to determine the effect on the DV
What are RCTs?
- highest level of evidence
- reduced bias
- ran done allocation of pts to intervene group or control/usual care group
- use CONSORT guidelines
What is the consort statement?
- consolidates standards of reporting trials
- 25 item checklist for elements of RCT
- flow chart of participant progress
What is a control group?
- group of participants whose performance on the DV is used to evaluate the performance of the experimental group on the same DV
- without a control group it is not possible to know if changes in the DV are due to the IV
How do you reduce the possibility of extraneous factors having an impact on the outcome?
- manipulation of the IV
- comparison/control group
- randomisation
What does randomisation ensure?
- group assignment is independent of participant characteristics
- every person in a target population should have equal chance of being selected
- every selected person should have equal chance of assignment to experiment or control group
What are the different types of randomisation?
- simple
- block
- stratified
What is simple randomisation?
Computer generated random numbers
What is block randomisation?
Ensures that the numbers of participant assigned to each group is equally distributed; commonly used in smaller trails
What is stratified randomisation?
Ensures that important baseline variables (potential confounding factors) are evenly distributed between groups
What are the elements of randomisation before an assignment?
- sequence generation
- allocation concealment
What are the elements of randomisation after an assignment?
Blinding
What is sequence generation?
Rule for allocating interventions should be based on chance
What are adequate methods of sequence generation?
- dice
- shuffled cards
- random numbers
What are inadequate methods of sequence generation?
Systematic (predictable) methods e.g. Date of birth
What is allocation concealment?
- works in conjunction with sequence generation
- shields those enrolling participants from knowing upcoming assignments
What are adequate methods for allocation concealment?
- numbered
- opaque
- sealed envelopes
- sequentially numbered drug containers
- central randomisation
What is blinding?
- could participants/study personnel have knowledge of group assignment, if so, could the knowledge affect the outcomes?
What is the role of the p-value when testing a hypothesis?
Reports the the probability of obtaining a study result of the null hypothesis is true
What does a small p-value indicate?
- results are unlikely when the null hypothesis is true
- null hypothesis is rejected
(
What does a large p-value indicate?
- results obtained are likely when the null hypothesis is true
- null hypothesis is accepted
(>0.05)
What is type 1 error?
Finding a significant difference when it doesn’t exist and rejecting the null hypothesis
What is type 2 error?
Failure to find a significant difference when there really is one and accepting the null hypothesis
What are the characteristics of true/classical experiments?
- randomisation
- manipulation
- control
What are the constraints of a true experiment?
- in a real life setting, controls are difficult to apply
- individual nature of human subjects
- expensive
- difficulties with randomisation
What are the different sources of bias?
- selection bias
- performance bias
- attrition bias
- measurement bias
- Hawthorne bias
How do you minimise selection bias?
Random assignment to groups
How do minimise performance bias?
Mask/blind interventions if possible and clearly defined protocols
How do minimise attrition bias?
Follow up all who leave the study and faithful reporting of attrition
How do you minimise measurement bias?
- reliable, valid measurement tools/instruments and trained testers/data collectors
How do you minimise the Hawthrone effect?
Compare control and experimental interventions
What are the difficulties with experimental designs?
- manipulation of IV or randomisation may not be feasible/ethical
- DV may not be measurable
- long term or rare outcomes or those unkown to be important at start of trial
What are the elements of a research question?
- PICO
- patients/population/problem
- intervention
- comparison
- outcome
