Selecting study design Flashcards
Why is randomisation important?
Key component of a robust statistical analysis how ever isn’t always completely possible
What are different types of bias?
Selection, Recall, Measurement, Survivorship, Responder ect
What is measurement bias?
Measuring something else and using it as an indicator of the thing we are interested in but cannot measure (proxy measures)
What is recall bias?
People recalling the past incorrectly eg breast cancer diet example
What is survivorship bias?
The sample only being those that have survived an ordeal eg the plane armour example
What is confounding?
When an observed relationship between two variables is caused by a joint relationship with a 3rd (confounding) variable
How can confounding’s impact be limited?
Random sampling
Matched sampling
Stratified sampling
Restricted sampling (eg one age group)
What is matched sampling?
In a case-control test pick one in each group that is effected by the confounding eg same age
What is stratified sampling?
Sample within confounding variable groups eg age
however means you end up with strata specific effect sizes
What is multiple regression?
Allows for multiple explanatory variables to modelled against response variable - lets confounders be controlled
What are pilot studies?
small scale study that aims to provide preliminary data to help plan a larger study
What is a descriptive study?
uses existing or easy to gather data - useful for hypothesis generation but poorly controlled
What is a correlational/observational study?
uses natural variation in observations to explore chosen hypothesis
What is a manipulative study?
artificially manipulated a system in order to explore hypothesis
What are studies of studies?
meta-analysis, systematic reviews
groups evidence to consolidate it
What is a cross-sectional study and why is it good?
where a random sample is taken from a single population at a given time
response & explanatory variables are measured at the same time
(cheap & easy)
What is a case-control study and why is it useful?
useful when response variable is binary (yes or no)
select individuals based on response variable and look for explanatory variables
What is a cohort study and why is it useful?
Individuals are selected according to exposure variable and tracked to measure impact of response variable
What is clinical study and why is it useful?
gold standard of medical intervention study
A manipulative study where individuals are recruited with pre-set criteria one group given treatment the other placebo
Randomised
Double-blind - researchers & patients don’t know what treatment is being used
Examples of 4 different studies that can be better than large-scale randomised sampling in certain situations
(Pseudo-replication)
Multiple replicates from individuals
Longitudinal studies - over time
Spatial Studies - spatial patterns
Cluster-sampling - replicate within sample
What do systematic reviews aim to do?
produce a quantitative or qualitative review of all the available literature relating to a hypothesis of interest
What do Meta analyses aim to do?
Use statistical method to combine quantitative estimates of outcomes across multiple studies
What do systematic reviews and meta analysis need to avoid?
Publication bias
hard to get replicate or negative result studies published
