Sampling, Replication & Between-Individual sampling Flashcards
Why is sampling needed?
Because counting an entire population is impossible and larger studies take too long/ too much money
Sample definition
a portion of the population (but represents the population)
- Sample replication helps find the interesting patterns of variation)
How can a small change in a mean results in such a large increase in extremes
because the whole distribution graph shifts (includes much more extreme at one end now)
What can errors bars show? (3)
Standard deviation
Standard error of mean
95% Confidence intervals
How is standard error of mean calculated?
Standard deviation / square root of sample size
What does standard error of mean show and how does increasing sample size effect it?
An estimate of precision of mean
Larger sample = small SEM error bars (more precise)
How do SEM & 95% confidence intervals compare?
95% CI are much wider than SD (both decrease as sample size increases)
How can P-values be misleading?
A very small p-value can indicate little effect (common with large sample sizes and no effect size comment)
What are the limitations of random sampling?
it wont capture some rarer groups (not all random samples are representative)
What is stratified sampling?
Dividing members of population into homogenous groups and then sample the groups independantly
What rules should the choosing the strata follow?
- Everybody in sample should be in a group
- No groups should have overlapping characteristics
- Simple random sample applied with each stratum afterwards
What is cluster sampling?
involves choosing natural clusters of individuals eg schools when looking at children
What must you be careful amount when using cluster sampling?
The cluster (eg school) being a confounding factor
what about individuals not included in clusters?
What is convenience sampling?
sampling that is convenient (eg interviewing on the street)
What must you be careful amount when using convenience sampling?
Avoiding Selection bias - eg assumptions of age/race
Avoid Recruitment bias - think bout who would be happy to be sampled eg somebody with strong feelings about either side of question
Think about external validity
