The article explores how to test if a statistical model fits real data well, even when some details are unknown. The researchers use a method called bootstrapping to estimate the distribution of the test results. They show that this method works consistently for different types of tests, both for single numbers and for groups of numbers.