Statistics 101: Nature nurture nonsense

reposted, with changes, from dailyKos

Over at daily Kos, feemus wrote a very good diary about the Bell Curve (the book, that is).  That led to a discussion of nature and nurture.  All such discussions are silly.  To see why, go below the fold.  (Oh, and the Bell Curve is nonsense masquerading as science, as feemus and nearly all the commentators knew)

Take a human trait.  Almost any human trait.  Some of that trait is almost certainly caused by nature – by one’s genes.  Some of that trait is almost certainly caused by nurture – by one’s environment. 

Let’s take a trait that we understand well: Phenylketonuria.  It’s 100% environmental AND 100% genetic.  How’s that again?  Well, for details, see the wiki page.  But for those who don’t want details, it’s a disorder that is caused by a defect in a chromosone that leads to an absence of a certain enzyme.  As a result, the body can’t metabolize a certain amino acid and there are dire consequences, including early death. 

OK…..it’s all nature.  If you have this defect, you have the disease.

But….if you avoid the amino acid that your body can’t process, there are no symptoms.  So, it’s 100% environment.

The reason for such nonsensical statements being nonetheless true is that, in PKU, as in many traits, genes and environment interact.  That’s a statistical term.  I’ll explain.

Suppose you have one variable – we’ll call it DV – (here, dying from PKU) that is affected by two other variables – we’ll call them IVs –  (here, having the faulty chromosone and eating the amino acid).  An interaction occurs when the effect of one IV on the DV is different at different levels of the other IV.  Here, if you don’t have the gene, the food doesn’t matter, and if you don’t have the food, the gene doesn’t matter.

When there is an interaction, the main effects are meaningless on their own.  What’s the effect of the food on dying?  Impossible to say.  It depends on the gene.  What’s the effect of the gene on dying? Impossible to say, it depends on the food.

OK, now let’s take another trait, one we understand less well.  Let’s take a personality trait like being a bully.  I haven’t done any research on bullying, but I’d bet that there are genetic causes. And I’d bet that there are environmental ones.  And I’d bet they interact.  Level of adrenaline is probably related to bullying behavior, and that is, in turn, partly caused by genetic factors.  But I’d be stunned if parenting didn’t affect bullying, and I’d be stunned if other environmental factors didn’t also affect it.  How might an interaction work?

Well, the effect of parenting on bullying probably depends on the personality of the child.  If the parent and the child ‘match’ in some sense, all may be well.  But the same parenting style with a different child might be terrible.  A father who is, say, a former marine who is into football and Nascar might be a great match for a child who likes similar things, but might have trouble with a child who is interested in art and poetry and hates physical activity.  That could lead to bullying.

So, that’s ONE reason nature nurture is nonsense.

There are others.

To determine how much of something is something, we need to be able to MEASUREE the things.  If we say, for example, that 40% of women are under 5’6″ (I have no idea) then we can get a bunch of women and measure them and see if we are right.  If we say that each inch of height in an adult male human is related to 3 pounds more weight (I have no idea) we can test that by measuring males’ heights and weights.  We know how to measure height and weight.

But…..if we take bullying.  Well, we don’t really know how to measure it very well.  And we don’t know how to measure environment at all.  And we don’t REALLY know how to measure the genes (single genes – fine.  Most human traits depend on lots of genes, and we rarely know which ones).

So we are saying that there is a relatinship between X, Y and Z, when we don’t know how to measure X or Y or Z.  Hmmmmm…… we might be able to say that there IS a relationship.  But determining the relative importance of X and Y on Z is not possible.

There’s YET ANOTHER reason why it’s nonsense.  Take the studies of twins reared apart.  Then you can say that there is NO genetic difference.  So, any difference in the trait MUST be due to environment.  But….well, sorry, it’s more problematic.  First, identical twins reared TOGETHER aren’t the same (damn humans messing up these nice theories).  Second, we don’t know how to tell how different the different environments are.  When children are raised by people other than their biological parents, it if often by people who are similar in various ways to their parents.  But how do we tell how different two homes are? We don’t know what to measure!

Nature is important
Nurture is important

The rest is nonsense