Also very true. Unfortunately this data is dirty, context dependent, or plain missing. This is true of pretty much any "real" experiment.