Control Groups for Breakfast, Revisited 

 A few months ago, I wrote an entry entitled   The Value of Control Groups in Causal Inference (and Breakfast Cereal) .  It was a report on a fun experiment I did that worked well both in my daughter's kindergarten class and my graduate methods class at Harvard.  There were a fair number of comments posted in the blog, and I also received dozens of other notes from parents and school teachers all over the country with many interesting questions and suggestions. 

 That correspondence covered four main points:  
 


  
  Some people suggested a variety of interesting alternative experiments, which is great, but in designing these many forgot that you must always have a control group.  That's the main lesson of the experiment: you often learn nothing without some kind of control group, and teaching this to kids (and graduate students!) is quite important. 

  Some people didn't squish the cereal enough and the magnet didn't pick up the pieces.  It will attract only when squished very well since the bits of iron are very small.  

  People then asked why the cereal doesn't stick to the magnet  without  squishing it up.  The reason is the same reason a magnet won't pick up a nail driven into a log, but it will pick up the nail if not in the log.  

   Finally, most people asked for other experiments they could run with their kids.  For that, which I'm writing up now, please tune in next time!