Again, I stress that none of this goes against your central thesis about relative spending. I think that the optometry finding actually supports what you are trying to argue, but not for the reasons you gave. I think we can say that the strongest evidence of benefit was for one of the cheapest and least charismatic parts of medical care: basic optometry. Following this logically, we can say that this is the sort of medical intervention for which there is good evidence of benefit from spending. Instead of excluding it from your analysis, you should be including it as the sort of medicine that should be getting more funding instead of the miracle-cure stories that get all the media attention even though ninety percent never make it to market.

I agree with your premise: the US spends a fortune on health and has very little to show for it compared to other OECD nations. David Cutler and Dana Goldman also agree. In fact, I think one would be hard-pressed to find anyone in health economics or public health who disagrees, although I’m sure there are plenty of people ready to be groomed for a Bjorn Lomborg role in health economics should there be sufficient threat to the medical industry.

But I disagree that the other economists accept your reading of the findings. Both Cutler and Goldman discuss their disagreements with your interpretation of the RAND study. These are not major disagreements, to be sure, and they are a great deal more measured in their responses than I was. But then, their interest is in health economics, whereas my interest here is the use of statistics in medicine. Actually, I do have some disagreements with the economic side of your argument, but these are not in my area of knowledge and they did not interest me as much as the statistics so I left them out entirely. It may be that the other health economists have set aside objections they had about the stats in order to address the economics.

Of course there is always room for disagreement about complex issues like the distribution of health spending. But in the use of statistics, to be frank, your article was wrong. There is no other way to put it. You excluded the best result by reclassifying it as “physics” rather than medicine; you repeatedly referred to findings in the subgroups in the the study populations that were least likely to benefit while not reporting data for the entire group; and you rejected statistically significant findings that didn’t suit your argument while enthusiastically accepting findings that were not statistically significant but did suit your argument. These expeditions go well beyond the boundary of differing opinion and extend well into the land of frank statistical error. Here Be Dragons.

This does not mean that your conclusions are wrong. It is quite possible to use faulty statistical reasoning and still arrive at a correct conclusion; this happens all the time in medical research, much to my perpetual irritation. I would not suggest that the statistical errors in your article are reason to dismiss your premise. I try to encourage medical students not to dismiss findings out of hand just because the study behind them is flawed. Every study is flawed to some degree.

But even when I agree with an opinion, I get rather agitated when I see it supported with bad statistical arguments, regardless of whether they are rhetorically useful. Statistics is a widely abused science, and as a result it is easy for people to dismiss results they don’t like with Disraeli’s “lies, damned lies and statistics.” I don’t expect to change the statistical contortionism of anti-rationalists like creationists and global warming deniers, but there is a large pool of intelligent, interested readers who deserve to be given the best possible information, and in medical research that necessarily means statistics. Cruddy statistical arguments, even in support of good causes, poison the well for all statistics.

But as I am now married to an associate professor, I feel compelled to rise to her defense. Not *all* of them are hopeless dabblers. :-)