To reply to Tristan, I confess that I hadn’t picked up that particular paper. If that qualifies as a debacle scale oversight then I must say I respect the high standards you hold.

With regard to the paper itself, I have to say that it doesn’t change the conclusions I came to. I spent some time familiarising myself with multiple linear regression a few years ago in order to solve problems similar to the one that paper tries to investigate. I started with high hopes for it as a technique, but I found it had many drawbacks. In particular the way you appear to be able to quantify the effect of different factors is beguiling. The trouble is you generate some data and then apply the technique, and you get some numbers. There is a great temptation to think you have something meaningful when in fact you don’t. This is particularly problematical if you look at lots of factors at the same time and ignore interactions between the different factors. In that paper they did not find any association between smoking and skin wrinkling. I find that very hard to believe, so I feel quite entitled to reject their conclusion that meat eating is factor that leads to skin wrinkles.

But thanks again for drawing this paper to my attention. I enjoyed reading it – I may even do a blog post on it when I have some time.

Bloggers (especially ones that haven’t done any professional writing) rely on the conversation to help keep them in line when or if they’ve made glaring errors.