A common but misleading statistical phenomenon rooted in the differing sizes of subgroups, where combined data hides the true trend. Simpson's Paradox reveals that aggregated data can appear to reverse important trends in the numbers being combined.
Simpson's Paradox is often an unintentional result of analysis when data from subgroups is incorrectly aggregated to represent larger trends. Examples have been found in many popular studies, such as unemployment rates by level of eductation, airline delay rates by airport, and even kidney stone treatment rates by procedure.
In the Lean world, there have been many pull system implementations that have failed because they were sized based on weekly or monthly buckets of demand, artificially dampening out the trends and true variation of the system.
December 2, 2009 Wall Street Journal Article - http://online.wsj.com/article/SB125970744553071829.html