Power & Sample Size Flashcards
Power definition
probability of correctly rejecting the null hypothesis if it is false
type 2 error
false negative; not rejecting the null hypothesis when it is false
type 1 error
false positive; probability of rejecting null hypothesis when it is true
which error relates to power
type 2 error
>power = 1-P(type 2 error)
which error relates to significance level
type 1 error
lower significance will require….
more patients
minimum clinical difference relates to…
effect size
minimum clinical difference definition
the difference you would want to see to conclude the treatment has had an effect
sample size test for continuous outcomes
difference in means
>common SD (prior knowledge)
>based on t-test
>data can be non-parametric
what is d
the effect size; mean difference/SD
sample size test for binary outcomes
difference in proportions
>based on chi-squared or fishers test
>variance of absolute difference with sample size
dropout adjustment
> usually 20% but based on previous studies
i.e new sample size = n/0.8
cluster trials (sample size calc.)
> need to account for correlation between individuals within clusters
inflate usual sample size by design effect
