Almost everything you read or see in the press or the media that makes use of statistics is at the minimum wrong and misleading, and at least half is intentionally fraudulent. The fraud is always of the same sort, namely efforts by the unholy alliance of academics looking to build up their visibility and funding with government officials looking to order people around and grow their bureaucracies and budgets.
This works because most people are very naive when it comes to statistics. Admittedly, it's a complicated subject, and you can't know everything. I majored in math, and took a course in statistics, but I'm certainly not an expert in statistics. However, I do know the one thing of significance. This is an extremely important thing to know, so I'm putting it in bold: From an observational or epidemiological study (as opposed to a random double-blind study), Relative Risk of less than about 3 is meaningless, not statistically significant, and no conclusion can be drawn from it. If you can learn that one principle and internalize it, it will be a real eye-opener in your reading of every day's news.
From the department of "I told you so," let us take up the subject of salt. I have been saying for years that salt in the diet is not a problem and you should eat as much as tastes good. Why did I say that? Because I have been looking to see if there is any observational study showing a RR over 2, or a significant random double blind result, and I've never seen it. QED. Obviously and as usual, I have all the forces of the great and good know-it-alls arrayed against me. For example, the Federal government gets into the act through the National Dietary Guidelines, issued jointly by the Departments of Agriculture and HHS. The World Health Organization of course piles on. The American Heart Association has totally bought into the anti-salt jihad. Mayor Bloomberg of New York has somehow become a national leader on this subject, and just a few months ago, according to the New York Times on February 13, "declared victory" in his national anti-salt campaign by strong-arming 21 food companies (including Kraft, Goya and FreshDirect) into lowering the salt content of their foods.
What evidence have these people had for their campaign? Essentially nothing -- some very weak observational studies and, according to yesterday's New York Times, "the well-known fact that blood pressure can drop slightly when people eat less salt [together with] other studies linking blood pressure to risks of heart attacks and strokes [and] models showing how many lives could be saved . . . ."
Pushing back against this ignorance, we have had only a few lonely unfunded amateur individual voices. On the salt issue, my favorite is John Brignell of the website "numberwatch." At a post from 2008 called March of the Zealots, he says this about the salt campaign:
Weirdest of all, but so typical, is the anti-salt campaign. It seems to have no other function than to keep the names of certain professors in the newspapers. The paucity of the evidence offered in contrast to the drama of the claims and the draconian nature of the demanded action is quite startling, but so characteristic of the genre.
Brignell also has this brief write-up to help you understand the concept of Relative Risk. A couple of useful quotes:
"As a general rule of thumb, we are looking for a relative risk of 3 or more before accepting a paper for publication." - Marcia Angell, editor of the New England Journal of Medicine"
"My basic rule is if the relative risk isn't at least 3 or 4, forget it." - Robert Temple, director of drug evaluation at the Food and Drug Administration.
Well, the salt jihad has always been such complete BS that it was only a matter of time before someone took a look at some real data. A couple of days ago, the Institute of Medicine, at the behest of the Centers for Disease Control issued a report examining the existing evidence. From the article in yesterday's Times:
In a report that undercuts years of public health warnings, a prestigious group convened by the government says there is no good reason based on health outcomes for many Americans to drive their sodium consumption down to the very low levels recommended in national dietary guidelines.
It's actually worse than that, because some of the recent studies cited actually found adverse health consequences from achieving low levels of salt in the system.
“As you go below the 2,300 [milligrams of salt per day] mark, there is an absence of data in terms of benefit and there begin to be suggestions in subgroup populations about potential harms,” said Dr. Brian L. Strom, chairman of the committee and a professor of public health at the University of Pennsylvania. He explained that the possible harms included increased rates of heart attacks and an increased risk of death.
You mean that the anti-salt campaign may actually be killing people? The incredible thing is that the campaigners seem never even to have thought of that possibility.
OK, you may be thinking, it's stupid, but what's the harm? You can still get salt and salt up your food if you want (the options aren't quite so good if you are Kraft or Goya). But salt is only one example out of hundreds of this kind of nonsense, so let's turn to a really important one, the standard for DUI. Turns out that yesterday's NYT also had an article on that subject. Reason: on Tuesday the National Transportation Safety Board issued a recommendation that the states reduce the allowable blood-alcohol concentration from 0.08% to 0.05%. It's the difference between basically 4 glasses of wine for a man and 3 for a woman putting you over, to 3 and 2. In other words, it is a proposal to criminalize essentially half the population. What is the authority of the Federal government to do this? The answer is that starting with a bill signed by Clinton in 2000 they coerced all the states that previously had a 0.1% standard to go down to 0.08% by threatening to withhold highway funding, and they could well do that again. This is very serious business.
So what is the evidence? Drum roll!!!!
People with a blood-alcohol level of 0.05 percent are 38 percent more likely to be involved in a crash than those who have not been drinking, according to government statistics. People with a blood-alcohol level of 0.08 percent are 169 percent more likely.
Those numbers sound rather paltry, but "169 percent more likely" (Relative Risk of 2.69) at least approaches the bare minimum of statistical significance, so where did the Times get the "government statistics" they refer to? They don't say, but a little searching on the internet turns up this collection of academic papers, the first one of which (Compton, et al, no date in the pdf) contains the numbers cited by the Times. Read the paper if you can stomach it, but it turns out that their raw data only came up with a relative risk of 1.57, and then they made a series of adjustments based on assumptions to get to the 2.69. For example, they made an assumption that there were a bunch of hit and runs that got excluded from the data because the driver was not available to be tested, and that that driver's blood alcohol must be elevated. In short, they made up the 2.69. The only data they have is 1.57, and there is good reason to think even that is elevated -- for example, the drinkers may be on average younger and thus less experienced drivers, or may have more propensity to drive with others who distract them. And by the way, the authors compare the Relative Risk from their data against the results of the previous major study from Grand Rapids in the 60s, which came up with a 1.88 Relative Risk at 0.08% blood alcohol.
Relative Risk of 1.57, or even 1.88? That is NOTHING!. For that matter, 2.69 is pretty damn close to nothing. We're talking here about criminalizing half the population! Now let's look at quotes in the Times from some of the creeps promoting more government control:
“These tragedies affect both sides,” said Robert J. Sumwalt, one of the board members. He said his wife’s first cousin was killed by a drunken driver who was traveling down a highway in the wrong direction. And his own cousin, he said, goes on Sundays to visit her 21-year-old daughter who was sentenced to 15 years in prison for drunken driving. Progress is mostly a matter of political will, board members said.
OK, Sumwalt, are you telling us that the guy who killed your wife's cousin had blood alcohol between 0.05 and 0.08? Of course not - if he knew the exact BAC and it were useful he would have said it. And believe me, nobody with BAC of 0.08% is so drunk as to drive down the highway in the wrong direction. This is just craven use of irrelevant tragedy in aid of a naked power grab for the government.
Well, people, I live in Manhattan, which is the only place in the USA where we don't have to drive home after drinking. If you are outside Manhattan, it's your issue not mine. But you had better wake up quickly.
UPDATE, 7:30 PM: Just now riding in the elevator here at 787 Seventh Avenue and the news screen has the headline "People who eat peppers have 50% lower risk of developing Parkinson's." In other words, yet another one of these phony studies that came out with no statistically significant results and yet is blared around the world as if it means something. Once you are on to this issue, you start to realize how nearly everything you read is a lie.