Rosa Brooks: Statistical Illiterate
Rosa Brooks consistently writes the weakest and most poorly reasoned columns in the L.A. Times — quite a feat, to be sure. She does not disappoint this week with her laughable piece titled The dark side of faith:
IT’S OFFICIAL: Too much religion may be a dangerous thing.
This is the implication of a study reported in the current issue of the Journal of Religion and Society, a publication of Creighton University’s Center for the Study of Religion. The study, by evolutionary scientist Gregory S. Paul, looks at the correlation between levels of “popular religiosity” and various “quantifiable societal health” indicators in 18 prosperous democracies, including the United States.
It’s official: Rosa Brooks does not know what she is talking about. The study says nothing whatsoever about whether “[t]oo much religion may be a dangerous thing.” Here is a quote from the study itself (my emphasis):
This is not an attempt to present a definitive study that establishes cause versus effect between religiosity, secularism and societal health.
That’s because correlation is not causation. This is a concept I have flogged on this site so many times that my readers are no doubt sick of it by now, but no matter how many times I repeat it, some people don’t seem to get it. I’ll say it again: correlation is not causation! Not even a little bit!
The funny part is that Ms. Brooks actually mouths those words — but then makes it abundantly clear that she doesn’t have a clue what they mean:
Although correlation is not causation, Paul’s study offers much food for thought. At a minimum, his findings suggest that contrary to popular belief, lack of religiosity does societies no particular harm.
Wrong, wrong, and wrong! Brooks is just statistically illiterate.
According to Brooks, correlation may not be causation, but it certainly suggests the absence of an inverse causative effect. In other words, she admits that the fact that we have more of X and more of Y may not show that X causes Y. But (she says) it certainly suggests — at a minimum! — that less of X will not have any effect on the number of Y.
That is just flatly false.
Let’s take an example to make this less abstract. As population increases, you will see more doctors. As population increases, you will also see more sickness, because there will be more people — sick people and well people. In other words, you will see a positive correlation between the number of doctors and the number of sick people.
Does that mean that reducing the number of doctors will have no effect on how much sickness there is in the world? Not at all — and in fact, common sense tells you that quite the opposite would be true. But if you substitute “doctors” for “religious sentiment” and “sickness” for “societal problems,” you could easily see Brooks writing the following:
IT’S OFFICIAL: Too many doctors may increase sickness.
This is the implication of a study showing a correlation between numbers of medical doctors and numbers of sick people in 18 prosperous democracies, including the United States.
Although correlation is not causation, the study offers much food for thought. At a minimum, the findings suggest that contrary to popular belief, eliminating doctors will not cause there to be more sickness.
Why does this woman have a weekly column?
It must get confusing for LA Times editors to show causation of their insinuation that education spending is decreasing, with their insinuation that test scores are declining. In fact, spending is increasing and scores are improving. On Aug 15, the state reported a 5% improvement in English test scores over a year ago, and a 4% improvement for mathematics test scores. Education spending (total, and per student) has increased every year for at least 6 years.
Shredstar (532850) — 10/1/2005 @ 4:33 pmA better question would be, “How is it we are still referring to garbage studies?” The study here is a prime example of psedo-scientific rubbish.
In this one study we have indeterminate indicators, statistical insufficiency, bias at every aspect of the “experiment” and a refusal to even consider a refutable hypothesis.
Anyone what to take a stab at defending it?
Paul Deignan (d2fd7b) — 10/1/2005 @ 5:58 pm“At a minimum, his findings suggest that contrary to popular belief, lack of religiosity does societies no particular harm.”
Actually, physical health is not the reason that we have “religiosity.” Even if it has no effect on our physical health, a lot of us believe that a lack of religiosity will have a particular harm to our spiritual health.
CPAguy (cfffbf) — 10/1/2005 @ 6:25 pmBut the point is that it is pure speculation for her to posit that a lack of religiosity will not affect the factors in the study (homicide rates, STDs, etc.)
Patterico (4e4b70) — 10/1/2005 @ 6:52 pm“Why does this woman have a weekly column?”
Well, could the answer be: affirmative action?
There just aren’t enough female columnists to go around, so the LAT lowered the standards so they could make the grade, diversity wise.
My fingers are crossed, of course. I wouldn’t want to get into hot water like that awful man on the radio who wants to abort babies.
Black Jack (ee9fe2) — 10/1/2005 @ 7:01 pmBlack Jack,
What man on the radio wants to abort babies? In case you didn’t realize it, abortion is homicide. You are killing a human being. There are times when homicide is justified. The number of justifiable abortions is a small fraction of the nearly 1,000,000 abortions a year currently performed. The man that I heard on the radio wanted to minimize the abortion of babies.
Charles D. Quarles (5d11c1) — 10/2/2005 @ 12:31 amWe need a special smiley for irony. It wouldn’t look right for someone to write:
but maybe a different symbol would serve. How about
to represent a deadpan delivery.
Doc Rampage (b7bb1a) — 10/2/2005 @ 1:04 amShe hedges so nicely you can’t actually pin her down with your critique. Religion “may” be dangerous, the study has “implications,” “correlation is not causation,” it all could be the other way around, it “suggests, etc. I think she knows full well that you can’t conclude much meaningful from the study, she just wants to get some jabs in as much as she can without having anything solid to work with.
Al. (cec513) — 10/2/2005 @ 4:42 amNuts. It turned my text smiley into an emoticon, thereby ruining the joke. Damn you WordPress!
Doc Rampage (b7bb1a) — 10/2/2005 @ 11:11 pmWhat are Ms. Brooks’s credentials? Statistics is notoriously the easiest sub-major in the math curriculum, yet as you suggested, she almost but not quite contradicts herself with quibbles.
The whole problem here is deconstruction. Ms. Brooks selects those statistics that tend to agree with her conclusions and omits others. She also conveniently forgets, as did the original Journal of Religion and Society article, that Evolution theory has never had scientific proof by way of an observation of man descending from a lower creature. Good science contains observable quantities. Extrapolations beyond man’s observations is an “Adult Game.” This terminology concerning the Big Bang was privately communicated to me by the great mathematical physicist Artur Rosenthal, whose seminal papers on the ergodic and quasi-ergodic theorems are cited to this day, when I challenged him on the fact that nobody had observed, let alone repeated, the Big Bang. His answer in his accented German around 1950: “Vell, it’s adult games.”
Creation/Anti-Evolution theory has no observable verification. But neither does Evolution theory. Therefore, in my view they both are metaphysical in nature.
Here’s my letter to the LA Times pointing out another fallacy.
To the Editor of the Los Angeles Times,
The opinion article “The Dark Side of Faith” by Rosa Brooks (Saturday, October 1, p. B.15) contains many errors of commission and omission in attempting to show that “Too much religion may be a dangerous thing.”
Correlation is not causality, as we learn in any good course in Statistics 101 and as Ms. Brooks informs us. So correlation can be deceptive indeed. After quoting this elementary proposition, Ms. Brooks should have stopped there instead of gong ahead and claiming that the statistics that she selectively marshaled from but one study from the Creighton University’s Journal of Religion and Society offered “much food for thought” especially when her statistics were highly selective.
She cites the statistic that infant mortality rates are higher in some unnamed red states in the South and Southwest and lowest in New England. But she neglects to recognize that New York City has one of the highest of all places in the nation, and that generally speaking statistics show that infant mortality is much better correlated with poverty. So the culprit here is poverty and not religion. A better title for her article is “The Dark Side of Poverty” and even better would be “The Dark Side of Poverty, Human Sacrifice, and Cannibalism” for the following reason.
The opening line of the article reads: “Too much religion may be a dangerous thing.” It should have added: “But too little religion may be a horrible thing.” Throughout her article Ms. Brooks rants against absolutistic religion. She neglects the fact that there was an important moral need for such a form of religion. Cannibalism and human sacrifice occur in many primitive cultures, even to this day. One of the basic purposes of absolutistic religion was to instill the fear of God in those who practiced human sacrifice and cannibalism, and thereby wean the people from these horrible practices. It is true that some cultures that have no strong religious tenets do not suffer from either cannibalism or human sacrifice, but for all we know these cultures never suffered from it and hence never had need for absolutistic religion. But can we now throw away absolutistic religion without some backsliding into cannibalism or human sacrifice even in our own culture, to say nothing of the primitive cultures that still practice these abominations and hence have dire need for absolutistic religion? This sort of question Ms. Brooks does not address even though it looms large as a future problem as opposed to a restricted snapshot of the present state of the world — the restrictive, snapshot mode of thinking being a frequent defect in deconstructionism and postmodernism.
Sincerely,
David Caplan (701aae) — 10/6/2005 @ 12:31 amDavid I. Caplan
247 SE 3rd Avenue
Delray Beach FL 33483
Tel.: 561–330–3269
I hate to break it to you,
but it is you who is statistically illiterate and reasons poorly.
Indeed, you are correct that correlation does not prove causation. The reason for that, however, is what you are missing, and is also why your critique is fatuous.
The reason correlation does not prove causation is that, because some things we may wish to study do not allow is to randomly divide groups into a control group and an experimental group. This holds for any complex question.
In other words, we cannot randomly assign some people to be religious in one country, and others to be non-religious in another country, and then measure the outcome in any pure fashion. BUT, that does not mean that we cannot draw meaningful conclusions from correlational data.
According to your fundamentalist logic, we could never find evidence that smoking causes cancer in humans, for instance, because we can’t randomly assign some people to smoke and others not to smoke. Yet, if we look at epidemiological data, and follow people who smoke and people who don’t over long periods, we will find without a doubt that those who smoke are very many times more likely to get cancer.
Of course, this does not prove that smoking causes cancer in the absolutist sense, but it strongly suggests it. The more data that accumulates to support this hypothesis, the stronger the hypothesis appears. It just so happens that it is generally accepted in the scientific community that smoking does cause cancer in humans, although all we have are correlational data to support that conclusion. Why? Because the accumulated weight of the evidence, including ruling out alternative explanations, is just that strong.
Your simpleminded reading would suggest that if dozens of witnesses (and a security camera) saw a man leaving a building with blood on his hands, and a 44 in them at the precise time someone was murdered with a 44, we couldn’t prove beyond a reasonable doubt that the person did it. Sure, maybe it was a coincidence, but try telling that to a jury. Enough coincidences add up to something close to certainty.
Now, to your critique. You disparage her conclusion, “At a minimum, his findings suggest that contrary to popular belief, lack of religiosity does societies no particular harm.”
There is nothing statistically incorrect about this whatsoever. She stated that the data SUGGEST that, not that they conclusively prove it. In other words, if someone looking at the mounds of data on smoking and cancer concluded that, someone with your quite silly position would cry, “Foul! You can’t prove that! It could be that people who smoke also eat more ham, and that ham causes cancer!”
True enough; that COULD be the case. But still, the data SUGGEST that smoking causes cancer, just like the data here SUGGEST that a lack of religiosity does a society no harm. After all, if there were a causative influence between religiosity and social well-being (as religionists typically argue), we would expect to see better outcomes in more religious societies, just like we would expect to see lower cancer rates in non-smoking societies. You are correct that this would not PROVE a causal link, but it certainly suggests one. Therefore, your criticism is completely unfounded. You right wingers just like to pound square pegs into round holes (Iraq anyone?), and will take potshots at anything you can when they don’t support your predetermined conclusions.
Appropriately enough, this is what faith is all about, and I think your response SUPPORTS her reading of the data (but it doesn’t prove it, so just breathe deeply for a moment and relax).
So if you require any more lessons in logic or statistics, feel free to contact me. It would certainly be prudent before you go making yourself look so silly by criticizing a perfectly sound conclusion. Hell, if you get good enough at logic, you might even stop voting for right wingers.
Defenestrator (62067a) — 10/8/2005 @ 8:16 pmDefenestrator,
According to your logic, the data certainly SUGGESTS that tipping high on a restaurant bill causes lung cancer. Talk to any waitress, and she’ll tell you that she loves nothing better than a table full of smokers. That’s because they tend to be drinkers — running up the tab, and tipping high as a result of the high and their good feelings from the drink.
Since cigarette smoking causes cancer, you’ll find a correlation between high restaurant tips and lung cancer — guaranteed. If I’m right about that, and I’m confident I am, does the data SUGGEST that tipping high causes cancer?
According to you, it doesn’t prove it, but it certainly SUGGESTS it.
According to me, it doesn’t suggest (or even SUGGEST) a damn thing.
Or, to take the example in my post which you completely failed to respond to: would a correlation between the number of doctors and the number of sick people SUGGEST that eliminating doctors will not cause there to be more sickness?
Epidemiology does not base conclusions on the connection between smoking and lung cancer on simple correlations, as you suggest. They look at several other factors, including, quite importantly, possible confounding variables (or, more precisely, the lack thereof — by ruling them out). The study Brooks looked at explicitly disclaimed any attempt to do so. Therefore, the findings did not suggest a damn thing, your ridicule notwithstanding.
Tone it down and respond to the arguments. You don’t need the capital letters, just logic.
Patterico (4e4b70) — 10/8/2005 @ 8:53 pmIt’s true, Rosa is an elitist snob…
I once wrote to Rosa stating that: “A great 1/2 American once said: “When the eagles are silent, the parrots begin to jabber.” His name was W. Churchill.
She wrote me back saying: “…with respect, if you don’t like my columns, stop reading them! It will save you a lot of time.”
As far as I can tell, Rosa only wants anti-America activists, Atheists, Socialist & Communist sympathizers, ACLU activists, “Internationalists”, and turd-world radicals to read her columns. Bush-bashers are welcome too. She must think people in and from L.A. are too stupid to see past her agenda.
I’ll give her the point that the “editors” of the LA Times are not too stupid.
M. Emley (04ea4d) — 10/27/2005 @ 1:37 pmBrooks on Bush…
In the LA Times, Rosa Brooks misrepresents the Hamdan holding in order to score political points against Bush. Brooks writes:… the real blockbuster in the Hamdan decision is the court’s holding that Common Article 3 of the Geneva Convention applies…
ProfessorBainbridge.com (72c8fd) — 6/30/2006 @ 3:15 pmSir- I will not discuss the scientific validity of the claims made either by yourself or by Ms. Brooks. She might not be a statistician, but sorry to say, neither are you. The example you used about doctors and sick people both increasing with the increase of population show either you lack of knowledge about statistics or that you are trying to manipulate facts.
Tammam Aloudat (382c28) — 9/6/2006 @ 5:16 amHere it goes, in statistics, we DO NOT take absolute numbers. You use other variables such as MEANS and PROPORTIONS. So in your example, you should use the propotion of doctors to population and the proportion of sick people to population, both will be exactly the same (in the absence of other factors) with the incease of population. So in your poor example, there is no correlation to start with and to talk about possible causation.
Another point, correlation is not causation, that is true, but to extablish causation, you will absolutely need to establish correlation first. There are some criterias for going from correlation to causation, one of them is logical, physical, or biological relation… So we have established correlation between, say, HIV/AIDS and some sorts of fungal infections. We go then to establishing the causation and text the hypothesis by NON-STATISTICAL methods, in my example biological and laboratory means. Now in the case of religious and social issues, it is absolutely acceptable to establish the correlation statistically and then the causation by means of logic, sociological, psychological, philosophical, anthropological or ethnographic studies.