I stumbled across this letter to the editor the other week in The Age…
Statistics don’t tell whole tale
“A TEAM from Melbourne’s Murdoch Children’s Research Institute says its study, which has tracked 1520 young people’s drinking habits over more than 10 years … shows there is no safe or sensible level of drinking for adolescents” (12 April 2009, The Age).
No, it does not.
The people at the institute have performed a statistical study. Statistics can be useful, but they do not prove cause and effect. You can look at the results and say that starting drinking early leads to problems in later life; you can look at the same results and say that people who have the sort of personalities that lead to problems with alcohol often start drinking when they are younger.
The only reason someone would prefer one conclusion to the other would be preconceived ideas being projected on to the results, rather than letting the data speak for itself.
This becomes clearer in a later quotation from the lead researcher, epidemiologist Elya Moore. She said the trend was not as clear for girls, but that this “could have been a failure of the research method, rather than evidence that girls were more able to drink at safe levels”.
We are asked to doubt the results that do not support the answer the researchers were expecting, although these results were part of the same study that produced the good results.
– Lee Bradshaw, Carlton
Bingo. It’s always nice when people like Lee Bradshaw from Carlton “just get it”. Researchers must always be mindful of the points that Lee has raised in that letter. Don’t confuse correlation with causation, and don’t let your own personal biases get in the way of a good study.