how to estimate sample size binary Flashcards
define estimation
estimation of effect (e.g. the difference in response rate between two groups)
define hypothesis
how unlikely you are to see an effect this big by chance if there is no genuine effect (null hypothesis)
is a large of small confidence better at predicting our estimate
small
smaller larger in which our try value is likely to sit
what 2 factors will determine a small confidence interval
less variation in dated
larger sample size
define power
Power is the probability of rejecting the null hypothesis when the null hypothesis is false
what factors make it easier to detect an association ( greater power)
- large mean effect
- variation in effect is smaller
- large sample size
what value is considered as significant for power
Usually the aim is 80% power i.e. probability of 0.8 of finding a difference/detecting an association
why would you calculate power before a study
- ensure study is well designed
justify your study to grant reviewers
why would you calculate power after a study
- demonstrate that you were sufficiently well powered.
- to estimate if there is an association,how small it must be for it to be missed.
how can be calculate power in RESS and why is this flawed
- estimate sample size needed
- does not account for variation
how are odds ratios different to risk ratios
always more extreme so the association ‘looks’ bigger
define odds of an event
the probability an event occurs / probability an event doesn’t occur
define odds ratio of 2 groups
odds of an event for ‘exposed’ individuals / odds of an event for ‘unexposed’ individuals
what values is a odds ratio if there is no effect
1
if an event occurs 20% of the time what are its odds ratio
the probability an event occurs / probability an event doesn’t occur
20%= 0.2
0.2/ 0/8= o,35
