[Myrmornis]

It's also my experience of research in biological sciences that it is a widespread belief/fact that in order to get published in a top journal, the analysis methods must be "fancy", for example involving sophisticated statistical techniques. I worked on computational statistical methods so I'm not against that per se, but the problem is that if you have the training to contribute at the research front of an area of biology you rarely have the training to understand the statistics. Some would say that the collaborative publication model is the solution to that, but in many cases the end result isn't what one would hope for. I do think that less emphasis on "fancy" statistics, and more emphasis on simple data visualizations and common sense analyses would be a good thing.

[FrojoS]

Agreed. And to add to that: The more fancy the statistics have to be, the less robust are the results.