Power Flashcards
What is a sample size of a statistical sample?
The number of observations that constitute it.
What is a sample size calculation (estimation) a measure of?
How many patients are needed in a study.
What do we use the sample of patients to draw inferences about?
The whole population.
What do Power calculations tell us?
How many patients are required in order to avoid making an “error” in our study.
Why do we need to calculate sample size?
- Need to know the necessary sample size for a study to have a high probability of finding a true difference if it exists. If it is too small we will lack the power to pick up that difference.
- Studies can be expensive - do not want to waste money on resources that will not be clinically valuable
- It is unethical to subject patients to treatment unnecessarily - either by performing a study with low power or by recruiting more individuals than necessary
What do the majority of statistical analysis involve?
Comparison, most obviously
between treatments or procedures or between groups or subjects.
When performing hypothesis tests we use the data collected
to find evidence to do what?
Reject the null hypothesis in favour of the alternative hypothesis
How do we test the null hypothesis?
Define the population
Take a random sample
Find a mean and standard deviation for both samples
Test the null using an independent samples t-test
When testing (trying to falsify) our Null Hypothesis what two types of error
can occur?
Type 1
Type B
What is power of a study?
The probability that the study will detect a predetermined difference in measurement between the two groups,
if it truly exists, given a pre-set value of alpha and a sample size N.
How can power be low?
The difference being sought is small
The sample is small
There is a great deal of variability in the data (standard deviation is large)
How certain we want to be in order to avoid a type I error
What does it mean if a study has low power?
Statistically significant differences will be hard to detect
What is a naive research question?
How many people do I need in order to find a difference in lung function measurements between smokers and non-smokers ?
What is an appropriate research question?
What is our appropriate sample size ,
if we want to have a 80% power to detect a statistically significant
difference of 0.5 litre in lung function measurements (FEV1) between
smokers and non smokers at a significance level of 5% ?
Choosing power, significance level and clinical difference involves what?
We set power to a desired level and then design the study to ensure that this level of power is achieved. Most studies seek a power of 80%-90%
We set our significance level ( usually at 5% ), that is the strength of the evidence that we require in order to reject the null hypothesis of no difference in risk
We set up our minimum clinical difference with the help of the previous data (relevant literature, pilot study)
We use the appropriate mathematical formula to calculate our sample size
