Tag Archives: causation

Correlation and causation: meditations on violence

XKCD looks at causation and correlation

XKCD looks at causation and correlation

With XKCD’s comic firmly in mind, I considered the news today. Constance Holden with the ScienceNOW Daily News over at Science Magazine drew my attention to an article linking violence to childhood sugar consumption. The telling quote is this:

Although lower education levels correlated with daily sweet-eating, the connection with violence remained significant even when the researchers controlled for factors such as family circumstances, parental attitudes, and IQ. “Try as I did, I couldn’t get rid of the sweets-violence connection,” says Morris.

So… it may be that highly sugared children (sugartots?) are made more violent by sugar… or it could be that violent people are drawn to sugar as children… the candy manufacturers may be drugging children to reduce their impulse control… or kids whose without parents failed to help them learn impulse control lack impulse control in adulthood… it’s all very complicated.

-Peter

Risk, statistics and ethics: the AIDS Vaccine

This idea of risk that we have been discussing on TBU for a while now has come up again. The results of a new HIV/AIDS vaccine study were released this week. The Thai trial has shown some promise. The incidence of HIV was 30% lower in the group vaccinated with RV 144 than the control group.

First: the basic bioethics question. Was it OK to give a lot of people a placebo which might let them think they were protected from HIV when they were not at all protected? (Answer: Yes) First-off, they were not told that they were getting an effective drug. They were told they were getting either a placebo or a probably ineffective, experimental vaccine. So the subjects knew beforehand that this shouldn’t be considered a real vaccine.

Second question: why not just give the vaccine to everyone in the study (16,000 people) and compare the effectiveness to the general population? The problem is that you would have a change in HIV incidence that was due to lots of factors. Behavior, knowledge, unknown risk factors (maybe people at higher risk had a greater desire to be in the study than the general population) all could affect the measured efficacy. How would you know which produced your result?

If the vaccine were 100% effective, then there would be no need for a placebo controlled trial. But nothing is 100% effective and – besides – how would you know before you tried?

Now, here’s the more difficult bioethics question: if you have a 30% effective vaccine, who should get it?

This is more tricky. You don’t want to encourage risky behavior (the ‘conservatives’ are always concerned about this). So there is a question to be answered by a careful psychology study: do people modify their behavior after receiving a drug that may or may not prevent a transmissible disease? It seems like they might, but scientists don’t make decisions on “might” if they can avoid it. We make decisions based on what is demonstrably consistent with experiment.

But then it becomes a quantitative statistics problem (more statistics!). It’s only worth vaccinating people if their behavior changes don’t outweigh the efficacy. And then it’s only worth vaccinating people who are at risk… but what if the higher risk people are more prone to behavior modifications? Is is possible to isolate a medium-risk category?

And in all of this, there are massive political problems, not the least of which are form the anti-vaccination people, which I will talk about next week. The vaccine is a real achievement, in any case. Lots of people thought it wouldn’t work. And it reminds us how complicated it gets when trying to do the right thing with imperfect tools.

-Peter