In the growing DogReader tradition of posting an article that very few people will find interesting (nor read), here is another statistical concept that the media mangle. This is happening with more regularity, as the pollsters turn out increasing data to chart the final stages of the American presidential election.
The error that the media keep repeating is assuming that there is a causal relationship to correctional data. Correlation does not mean causation.
Here is a nonsensical example to illustrate the point. Correlation is drawing together the possible relationship (or non relationship) between events. In this example, the correlation will examine drinking milk and alcoholism. Those people who have a problem with alcohol would be asked if, when they were children, they drank milk. The likely answer would be “yes”. The finding would be a high correlation – a relationship – that people who have a problem with alcohol drank milk when they were children.
The jump in logic would be to take these correlational data and imply causality. That would be that ‘drinking milk as a child will lead to alcoholism’.
The conclusion is obviously nonsensical. However, this is the type of causal relationship that the media presents as conclusions from data. It is, unfortunately, equally nonsensical.
Catherine Forsythe