Regression to the mean

2008-01-10

in Geek stuff, Science

Emily Horn at Canada Place

All manner of diets, supplements, and vitamins compete for customers and adherents. Given that we are creatures of biochemistry, it is plausible that such chemicals will have effects on human health. Unfortunately, they do not act alone, but rather within a complex web of interactions: genes, environmental effects, physiological changes, etc, etc, etc. This makes it exceedingly difficult to isolate and prove the effect of any particular substance, especially given that the effect in different people, or the same person at different times, may differ.

The phenomenon of regression to the mean is especially confounding when it comes to individuals. We do not generally change our regimen of diet or supplements unless we perceive something to be wrong. We take vitamins when we feel ill, and analgesics when we have a headache. Given that both illness and headaches tend to rise to a peak and then taper off naturally, virtually any action we take in response will precede an improvement. It doesn’t matter if you spend a pile of money on remedies, spend your time praying, or simply sit still and wait for improvement. Regression to the mean is the product of basic statistical mathematics and isn’t caused by anything chemical or biological. It is tautological to say that things are normally like the mean and that, most of the time, situations far away from the mean will be replaced by those closer to it. While it is true that we can do things likely to shorten or lengthen the period of illness or discomfort, it is virtually impossible for an individual to know whether such an effect has occurred. Did taking those vitamins shorten or lengthen the cold? Did it have no effect? What about those glasses of red wine?

The only sensible course of action is to essentially disregard our own experiences, except in such cases where there is both a reasonably large body of evidence (ideally in the form of a large number of double-blind and controlled trials) and there is a plausible explanation for the method of action. Failure to employ such checks against hasty false reasoning leave us vulnerable to the pernicious human tendency to see causal relationships everywhere, without the scepticism that is critical in separating conjecture from an investigated hypothesis.

Report a typo or inaccuracy

{ 6 comments… read them below or add one }

Milan January 7, 2008 at 4:17 pm

How to eat healthy? Eat food – real food
Toronto Star

“But we have almost no data on what those nutrients do in combination with other nutrients. We may be getting pretty close to a conclusion that it’s the lycopene in the tomato but, the truth is, it really may be other components in the tomato. Or the way it is reduced and simmered with olive oil and basil. We just don’t know. Only one thing is for sure: Don’t like the prevailing faddish nutrition advice? Wait five minutes; it’ll change.

That bit of news, of course, pretty much makes almost every nutritional supplement a massive waste of time and money. Yet Michael Pollan, in his new In Defense of Food: An Eater’s Manifesto, claims that people who take supplements are, for the large part, healthier than the rest. Quel paradoxe. Pollan doesn’t trust them either.
Given our hit-and-miss understanding of nutrients and what Pollan calls “reductionist” food science (reducing diets down to superfoods and those foods down to nutrients), he suspects that it’s simply because people who care enough to bother with supplements are probably also more careful with their diets in general.”

Mark January 10, 2008 at 2:38 pm

Unfortunately, rational thinking may not have high evolutionary fitness. If Frumpy the Squirrel eats some unfamiliar red berries and then gets sick, Frumpy probably doesn’t eat red berries again. Frumpy might have got sick for any number of reasons unconnected to the berries, but a squirrel that waits until it has statistically significant evidence before changing it’s behaviour is probably a dead squirrel. Better to turn up the gain on your learning, and accept that you develop some superstitions as a side effect.

For some related ideas about the value or otherwise of correct perception, see:
http://www.edge.org/q2008/q08_1.html#hoffman

Edward January 11, 2008 at 12:55 am

Thanks for writing this entry. I wish every patient would read your post before spending hideous amounts of money on pseudo-medical therapies that don’t work and might do harm. Many (but certainly not all) alternative therapies are simply well-marketed ploys to deceive the consumer.

. January 11, 2008 at 1:50 pm

The Basics: Introduction to Microbiology and Infectious Disease

Generally today, the all-encompassing group of “microbes” are divided into viruses, bacteria, fungi, and protozoans. I will discuss each category briefly, and touch on prion diseases as well, finishing up with a brief introduction to our body’s disease-control mechanism, the immune system. Please do keep in mind that the statements below are gross generalizations; exceptions exist to many of the basic guidelines put forth below. So to begin this veritable stew of information:

. October 16, 2009 at 10:52 am

Bracelets ‘useless’ in arthritis

Copper bracelets and magnetic wrist straps are useless for relieving pain in people with arthritis, say University of York researchers.

“People tend to buy them when they are in a lot of pain, then when the pain eases off over time they attribute this to the device.

“However, our findings suggest that such devices have no real advantage over placebo wrist straps that are not magnetic and do not contain copper.”

. May 2, 2010 at 7:00 pm

“The textbook talks in detail about therapies for different disorders including schizophrenia, anxiety disorders, and so on. The question which everyone is interested in is, “Does therapy work?” And this proves to be surprisingly difficult to tell. Part of the problem is if you ask people who go into therapy, “Did you get better after therapy?” for the most part they’ll tell you that they did but the problem is this could be a statistical byproduct of what’s called “regression to the mean.”

So, the idea looks like this. This line plots how you feel from great through okay to awful and it goes up and down and in fact in everyday life you’re going to–some days are going to be average, some days will be better than average, some days worse than average. You could plot your semester. You could do a plot every morning when you wake up or every night before you go to bed. You could put yourself on a graph and it’ll come out to some sort of wiggly thing. Statistically, if something is above average or below average it’s going to trend towards average just because that’s a statistical inevitability. When do people go to therapy? Well, they go to therapy when they’re feeling really crappy. They go to therapy when they’re feeling unusually bad. Even if therapy then has no effect at all, if it’s true that people’s moods tend to go up and down after you feel really bad you’ll probably improve rather than get worse. And so this could happen–the normal flow could happen just even if therapy has no effect at all.

When you talk about regression to the mean, it adopts certain assumptions. The assumption is there really is an average throughout much of your life and things go up and down within that average and for the most part that’s true for things like mood. For most of us, we have an average mood and we have bad days and we have good days. It’s always possible that you have a bad day and then from there on in it’s just going to go down and down and down but statistically the best bet is if you have a bad day you’re going to go back up to the mean. It’s–in some way you don’t even have to see it from a clinical point of view. You could map it out yourself. Map out your moods and the days where you’re most depressed sooner or later you’re likely to go up. Similarly, on the happiest day of your life odds are the next day you’re going to go down and there’s nothing magical about this. This is just because under the assumption that there really is an average in–built into one–each of us. If human behavior was arbitrary, it would be like a random walk but it’s not. We seem to have sort of set points and aspects of us that we fall back to that make the idea of a mean a psychologically plausible claim.”

Leave a Comment

You can use these HTML tags and attributes: <a href="" title=""> <abbr title=""> <acronym title=""> <b> <blockquote cite=""> <cite> <code> <del datetime=""> <em> <i> <q cite=""> <s> <strike> <strong>

{ 2 trackbacks }

Previous post:

Next post: