The Power of Statistical Power
In statistical hypothesis testing there are two hypotheses, two errors that can be made and the power of the test. Power is useful because having enough statistical power is necessary to draw accurate conclusions from the hypothesis test using sample data. The purpose of this article is to show how to easily it is to use power to determine if the correct hypothesis is being tested.
Additionally, it is always important to evaluate the errors in terms of the hypotheses and calculate what those errors are. For example, for lot release I believe the patient’s (or user’s or purchaser’s) risk is most important and should always be controlled and stipulated. This risk is the probability of releasing bad product or deciding an invalid measurement (assay) is valid. The manufacturer’s risk is also of interest. This risk is the probability of not releasing good product or rejecting valid results. This type of error costs manufacturers more money and can also impact drug supply. If only one statistical error is being defined, I feel it should be patient’s risk.
There are four parts to this article:
Part I: (A) Hypothesis testing and statistical errors
Part I: (B) How to use power to determine if the hypothesis test being used is correct