airing publication dirty laundry

I watched the full video of the recent International Journal of Epidemiology conference yesterday, which I was alerted to via Ben Goldacre’s blog.

It is over seven hours long so probably weekend viewing, but is well worth it. I particularly enjoyed Richard Smith’s assertion that journals now do more harm than good, Ben’s (different from pharma-bashing usual) presentation on transparency and why academia and the public sector are major transgressors, as well as George Davey-Smith’s themes that
(1) fancy models cannot solve issues of correlation and that
(2) experimental methods and, paraphrasing him “coming at the data from different directions, using methods that, even if worst case scenario are biased, will not all bias in the same direction”.

Indeed the latter is what my Sydney education drummed into me and which, sadly, a large proportion of the academic choice modelling discipline still doesn’t understand.

One of these days someone is going to try to replicate health DCEs and ascertain whether studies were done at a time when knowledge of key conceptual issues *cough*variance scale*cough* should have been known and applied. The results will be far uglier than what has been dug up regarding clinical trials.

Of course there are shining examples of good practice in academic choice modelling and I’ve had the pleasure and honour of working with them (both in the past and currently). I just feel there is something to a recently expressed view that academic reviews in certain areas are in fact worse than industry ones, due to a variety of factors, not least of which are incentives that too many people consider academics to be invulnerable to.