First the link to this week’s complete list as HTML and as PDF.
I fail to understand how Morton et al. could ever have been accepted for publication. Just because someone prefers something does not in the least imply he’s going to get it. Older women may well be – and are – less desirable, but they’re available and get pregnant all the same. Later children of older women have siblings to care for and look after them and thus probably a higher rate of survival and success than children of younger ones. A closer look at what the model does reveals the fallacy. Morton et al. do not model a preference (contrary to alpha males women can’t have more than one child every two or three years however preferred they may be anyway) but simply impose a strict rule of women not being allowed to mate at all after age 35 (Fig. 2b: “Bars under the figure indicate age classes allowed to mate.”). A silly assumption will generate silly results – GIGO.
With modeling and computer simulations becoming more and more important in different and diverse fields, constraining the sensible choice of parameters has attained key importance. As Zhang & Gong show, empirical fitting on the global outcome alone – the usual method for current modeling – does not suffice and obtains misleading results.
In my parents’ time university researchers had to build most of their experimental setup and instrumentation themselves and were highly aware of its capabilities and limits. In my time in the late eighties we had become accustomed to bought-in black boxes providing multi-digit readouts of something or other at the press of a button. With statistical packages for the evaluation and analysis of results things have become even worse now. My rule is and remains, that one must never use a program unless one could in principle achieve and understand the same result using pen and paper – albeit on a small and simplified sample. Joppa et al. give an overview of the extent of the problem.