How good are you at stats?

3 posts / 0 new
Last post
sichan
sichan's picture
How good are you at stats?

I was just reading a great paper on the over use of the Student's t-test in many biological studies.  The author surveyed a collection of studies from the journal "Behavioral Ecology" (published by Oxford University Press) and found that many utitlized some form of the Student's t-test.  This test should only be used when the variances of the two populations are identical (or nearly identical).  When this does not hold true, the Student's t-test actually does very poorly, as shown by the increased number of Type 1 and 2 errors.

I think as biologists, we often neglect to make sure that the tests we use are appropriate and that we need to stop treating these tests as one-size-fits-all-blackboxes.

The paper, entitled "The unequal variance t-test is an underused alternative to Student's t-test and the Mann-Whitney U test" by Graeme D. Ruxton.  The paper can be accessed here: http://beheco.oxfordjournals.org/cgi/content/full/17/4/688

Ivan Delgado
Ivan Delgado's picture
I could not agree more. It

I could not agree more. It has always been troubling to me how many scientists do not have a good handle of Math and Statistics principles. As someone that loves Math, but had to work hard at getting a good feel for statistics, I appreciate the value of having a good understanding of statistics, especially when designing your experiments. 
Thanks for sharing.
 

heehawmcduff
heehawmcduff's picture
It's my experience that

It's my experience that reviewers and journals are now placing more emphasis on appropriate and correct statistics (which can only be a good thing), although it still fairly apparent when dodgy stats do slip though the net in very high impact factor journals.