Comments on: Statistics, statistics http://timworstall.com/2009/11/21/statistics-statistics-5/ It is all obvious or trivial except... Sat, 11 Feb 2012 18:45:24 +0000 hourly 1 http://wordpress.org/?v=3.2.1 By: diogenes1960 http://timworstall.com/2009/11/21/statistics-statistics-5/comment-page-1/#comment-37527 diogenes1960 Sat, 21 Nov 2009 22:14:34 +0000 http://timworstall.com/?p=11289#comment-37527 hang on...that means that AGW becomes a meaningless political concept hang on…that means that AGW becomes a meaningless political concept

]]> By: Pat http://timworstall.com/2009/11/21/statistics-statistics-5/comment-page-1/#comment-37509 Pat Sat, 21 Nov 2009 12:48:00 +0000 http://timworstall.com/?p=11289#comment-37509 Amen, Nigel- a plug for JB is always a good idea. Amen, Nigel- a plug for JB is always a good idea.

]]> By: Nigel Sedgwick http://timworstall.com/2009/11/21/statistics-statistics-5/comment-page-1/#comment-37496 Nigel Sedgwick Sat, 21 Nov 2009 09:31:41 +0000 http://timworstall.com/?p=11289#comment-37496 Well said Tim. It's a very common problem: the quoting of the relative without the absolute. Of course, given that the right statistics do usually tell the truth, how else can opposing politicians each support their partisan cases? I've always been a great admirer of <a href="http://www.numberwatch.co.uk/number%20watch.htm" rel="nofollow">Professor John Brignell and his Numberwatch website</a>, which you too, of course, know about; but a reminder for the readership is no bad thing. There he gives us the <a href="http://www.numberwatch.co.uk/data_dredge.htm" rel="nofollow">data dredge</a>, in which scientists search one set of data for just about any possible 'trend', ignoring that some such 'trends' will occur by chance, if one looks hard enough. Thus a higher threshold should be set, for statistical relevance, when doing such a <strike>wide-scale investigation</strike> dredge. There is an equivalent 'mistake' by politicians. If you search hard enough for a narrow enough statistic in a perfectly decent overall analysis, you stand a good chance of finding one to support any side of the debate: cherry-picking. Best regards Well said Tim. It’s a very common problem: the quoting of the relative without the absolute.

Of course, given that the right statistics do usually tell the truth, how else can opposing politicians each support their partisan cases?

I’ve always been a great admirer of Professor John Brignell and his Numberwatch website, which you too, of course, know about; but a reminder for the readership is no bad thing.

There he gives us the data dredge, in which scientists search one set of data for just about any possible ‘trend’, ignoring that some such ‘trends’ will occur by chance, if one looks hard enough. Thus a higher threshold should be set, for statistical relevance, when doing such a wide-scale investigation dredge.

There is an equivalent ‘mistake’ by politicians. If you search hard enough for a narrow enough statistic in a perfectly decent overall analysis, you stand a good chance of finding one to support any side of the debate: cherry-picking.

Best regards

]]>