That thing I just linked to....might be bullshit.

As I always tell people before embarking on any kind of more complicated statistical analysis; first, look at the f**king data. Well, actually I'm a bit more professional than that, but that's the gist of it. The problem in this particular example is that they used as a control variable something that can be prior to the dependent variable-or not. The gender and fact of a child will inluence the decisions related to having another child. In order to really do this, you would have to factor in the effect of having a son or daughter as the first child, and then all of the combinations that arise with each additional child. Anyone who actually studied why people have children and what goes into that decision (or non-decision) would know that this analysis was bull. While that isn't my content area, I did spend a year on the NIH population and demography study section and THAT is the type of thing that they study. Sometimes an important part of statistics is simply good old-fashioned logic. If something seems too good to be true (or so weird that you know it will get published) it's probably due to something much more mundane.