Sidak reported an equation for estimating the probability value of statistical significance (p-value) required for individual tests that will allow for a target p-value to be achieved when all of the tests are used together.


Requirement: tests are independent of each other



(1) overall p-value desired (for statistical significance)

(2) number of tests to be used together


target p-value for each test =

= 1 - ((1 - (overall p value)) ^ (1 / (number of tests)))


If all of the tests have a p-value less than the target then the overall p-value will be less than its cutoff. In the real world a p-value might exceed the target if there was another test that would compensate for the change.


On the other hand, if there are multiple tests with a certain p-value, the final overall p-value if there are combined is:


overall p-value for a number of tests used together =

= 1 - ((1 - (p value for test))^(number of tests))


To read more or access our algorithms and calculators, please log in or register.