This is a statistical method that allows us to estimate the number of objects that must be studied in order to be able to statistically demonstrate a certain dependence. Basically, it can be stated that quite negligible and, from a practical standpoint, completely uninteresting dependences can be demonstrated in a sufficiently large set. For example, in order to demonstrate the effect of the shadowing of a field by flying swallows on the crop yield, all the arable land in Europe would probably have to be reserved for our “very important” study. Before a scientist decides to perform a more demanding study, he should first estimate, on the basis of the available information and using power analysis, how large a test set he will require to have a reasonable chance of demonstrating the studied effects. Power analysis simultaneously permits estimating post factum how far we can believe the results of a study that did not demonstrate the existence of the studied dependence.
- more or less305
- not at all628