Writing about web page http://www.resuscitationjournal.com/content/authorinfo
I submitted a manuscript to the journal Resuscitation recently. It’s a pretty well-regarded medical journal, with an impact factor (for 2013) of 3.96, so a publication there would be a good solid paper. While formatting the manuscript I had a look at the statistical section of the Instructions for Authors. This is what I found:
* Use nonparametric methods to compare groups when the distribution of the dependent variable is not normal.
* Use measures of uncertainty (e.g. confidence intervals) consistently.
* Report two-sided P values except when one-sided tests are required by study design (e.g., non-inferiority trials). Report P values larger than 0.01 to two decimal places, those between 0.01 and 0.001 to three decimal places; report P values smaller than 0.001 as P<0.001.
That’s it! 69 words (including the title), more than half of which (43) are about reporting of p-values. I really don’t think that many people would find this very useful (for example, what does “use measures of uncertainty consistently” mean?). Moreover, it seems to start from the premise that statistical analysis IS null hypothesis significance testing, and there are lots of reasons to take issue with that point of view. And finally (for now) it is questionable whether two-sided tests are usually the right thing to do, as we are usually interested in whether a treatment is better than another, not in whether it is different (not bothered whether it is better or worse) – won’t get further in to that now but suffice to say it is a live issue.
Original post 9 April 2015: http://blogs.warwick.ac.uk/simongates/entry/journal_statistical_instructions/