| "But in the presence of statistical evidence, don’t tolerate contrarian anecdotes, and don’t make them yourself, knowing the exaggerated impact they can have." When you start to design a statistical experiment, you have already made an important methodological choice. See http://en.wikipedia.org/wiki/Positivism http://en.wikipedia.org/wiki/Grounded_Theory I've noticed that in my own field which is education, there appears to be a fondness for sophisticated statistics, even though no manager ever allocated students to teachers on a double blind random basis. An excellent example is the way the UK Education ministry has decided that 'phonics' is the way to teach reading. http://www.bbc.co.uk/news/education-18462214 I think that this general tendency might be an example the 'white coat syndrome' in action; belief that using formal statistical techniques might increase the meaningfulness of the results. I suppose that is a form of cargo cult. This is hacker news, a forum aimed at people with novel business proposals and new software to try out. Should you be trying to find 'the Truth' or should you be building some grounded theory that tells you what to do next, provisionally, now, today? |