matthen

It can be hard for computers to find hidden causes for observations.  The above animation shows fictional statistics for the number of people going to the beach, ice cream sales, cinema ticket sales and umbrella sales.  The corresponding icon is bright if one of these statistics is above average on a certain day, otherwise it is dark.  From the data alone a computer might guess the model on the left, because everything is intercorrelated.  A bad statistician might then look at this and suppose for example that umbrella sales are caused by people not going to the beach. A better model is the one on the right, which introduces a ‘hidden variable’ not in the data- the weather.  It’s really the weather that is causing these things, and this gives a simpler model even though it produces exactly the same statistics. [code]