Pretty sure these correlations aren't statistically important. N events will give a deviation of ~sqrt(N) - which for N = 6000 is > 1%. Therefore seeing that 12

Pretty sure these correlations aren't statistically important. N events will give a deviation of ~sqrt(N) - which for N = 6000 is > 1%. Therefore seeing that 12% of Microsoft people buy Capri Sun but only 10% of Googlers is pretty meaningless. Furthermore, we don't know how many items the test was against, leading to only those items that had large discrepancies being reported. http://en.wikipedia.org/wiki/Publication_bias#The_file_drawer_effect http://en.wikipedia.org/wiki/Publication_bias#The_file_drawe...