Significance

This might be a good test to perform on all squishy research:  replace the numerical results with random numbers and see if the conclusions change.