Sampling Flashcards
Target population
Population to whom findings are generalisable
Study population
The proportion of the target population available to the researcher. Collected using a sampling frame.
(Representativeness, completeness)
Most commonly used sampling frames
- cancer registers
- GP lists
- professional registers (GMC, nursing & midwifery council)
- workplaces, schools, hospital clinics
What are the drawbacks of using the UK electoral register as a sampling frame?
Under-representation:
- young/students
- travellers
- homeless
Why do we use sampling?
Because investigating the entire population of interest is usually costly & impractical.
Name 3 methods of random sampling
- Simple random
- Cluster
- Stratified
Simple random sampling
From list of whole study population, draw a random sample. (ie- random number generator - equal probability of being chosen)
Cluster sampling
Random selection from within natural clusters in a population (schools, GP practices, households)
Often used in RCTs to reduce contamination btw patients
Stratified sampling
When measurement of interest is likely to vary btw different sub-groups, one may selectively recruit a higher proportion of some minority groups to increase statistical power and sample size
(equal chance of being picked but random higher proportion of one group)
Name 4 non-random sampling methods
- Systematic sampling
- Convenience sampling
- Snowball sampling
- Street survey
Interpretation - avoid inferences about gen pop
Systematic sampling
Pick every nth individual from a sampling frame (potential for bias if underlying patterns in sampling frame)
Convenience sampling
Choose study ptpts on basis of being readily available.
(Subject to volunteer bias)
Snowball sampling
Ask study ptpts to recommend acquaintances who meet pre-specified criteria
(Useful when studying hard-to-reach groups when no sampling frame is available)
Street survey
Pick ppl who appear to fall into certain categories of interest from the street & ask your Qs
(Often quota sampling - specify certain number M/F from certain age groups)
No sampling frame, difficult to know who findings applicable to