Support for Making Sense of Statistics
Two Liberal Democrat MPs have put down an Early Day Motion in the House of Commons welcoming Making Sense of Statistics, the guide published by Straight Statistics and Sense about Science, in collaboration with the Royal Statistical Society.
Julian Huppert, MP for Cambridge, and Bob Russell, MP for Colchester, say in the EDM: “This house recognises the importance of the appropriate use of statistics in policy-making and in public discourse; further recognises that parliamentarians and other policy-makers should be able to identify key statistical concepts, seek to avoid making obvious errors in interpretation and identify the key questions to ask when presented with data and statistical analysis”.
EDMs are seldom debated. They are a means of alerting members to issues. The two signatories have now been joined by nine more: Peter Bottomley, Mike Hancock, Stephen Williams, Kelvin Hopkins, Annette Brooke, Jeremy Corbyn, Paul Flynn, John Leech and George Andrew.
Making Sense of Statistics can be downloaded here. Printed copies are available from Straight Statistics and Sense about Science, free of charge.
Nigel Cook (not verified) wrote,
Sat, 17/07/2010 - 10:45
The danger is that formal statistics are used to mislead. The booklet states that the standard definition of statistically significant (when you use a Chi-quared test to compared expected with observed data) is that the probability of the results occurring by random chance is 5%.
So 1 in 20 "statistically significant" results that passes the Chi-quared test may be a fluke.
Even if the statistical correlation really is reliable, you then have the problem of interpreting it. Examples of misinterpretations with tragic results abound.
Darrell Huff's book "How to Lie with Statistics" gave one classic example in the 1950s, when researchers reportedly found a correlation between the number of children per family in Holland, and the number of storks nests on the roofs of the houses. The bigger the family, the greater the number of storks nests.
In politics, this "statistically significant" correlation would be used as "evidence" for whatever fanciful theory your imagination comes up with.
But actually, larger families on the average will tend to need bigger houses, and due to the expense will tend on average to have older houses. The larger the roof area and the older the roof, the more storks nests, so there is nothing mysterious going on here.
Post new comment