Monday, October 18, 2010

Research with one group


If we want to see if blondes have more fun, we could get 5 or 10 blondes and the same number of non-blondes, record their fun and see who has more.  But it may be cheaper and more informative to just get one group of blondes and record how much fun they have.  Then, after a while, darken their hair or give them non-blonde wigs and see how much fun they have then.  When we run an experiment by introducing an intervention such as wigs, it is important to look at the before-and-after picture.  The famous Campbell and Stanley book first alerted me to the meaning of time series experiments.  They present a chart like my simplified version below.  If the readings for fun show a pattern such as lines A or B, we can conclude that fun has been dropping in general and the wigs didn't disturb the fall.  But line C implies the darker hair dropped fun precipitously.  You can see how much more complex judging the effect would become with real data, such as US wheat production.  Again, the real world is complex and tricky.  Time and conditions matter.


Popular Posts

Follow @olderkirby