On 3 Jan 2001 15:07:31 -0800, [EMAIL PROTECTED] (Will Hopkins)
wrote:
> I'm a newcomer to understanding and calculating heritability and
> related statistics. I notice that heritability is a
> variance-explained statistic (variance attributed to inheritance
> divided by total between-subject variance). In the sphere of
> experimental research, variance explained conveys a false impression
> of the magnitude of an effect. You have to take the square root of
> the variance explained to convert it to a correlation coefficient,
> then interpret it using Cohen's scale of effect magnitudes (<0.1 =
> trivial, 0.1-0.3 = small, 0.3-0.5 = moderate, >0.5 = large). Thus, a
> variance explained of 0.01 (1%) is actually a small but non-trivial
> effect, because it is equivalent to an effect size of 0.1.
"Cohen's scale" is a set of numbers that works out pretty well for
social science research, or "behavioral sciences" as it says in the
title of his book. He is describing studies where the sample sizes
might be from 25 to 250 or so. I believe his book does contain
warnings about over-interpreting that scale.
If you have a criterion (reaction time, etc.) where you average dozens
or hundreds of observations to make a point to be analyzed, the
"effect size" is magnified by averaging. That is, if you can change
an average by .01, that fraction is a lot bigger fraction of the
between-Subject variance (of averages) than it is of the
between-trial variance.
What is 1% a fraction of? 1% might be a huge effect. I have seen
that described as the "effect size" for the difference between the
professional baseball player who earns millions per year, and the pro
who can't quite stay in the major leagues.
>
> So my question is this: should we take the square root of
> heritability to get an idea of the contribution of inheritance to a
> particular trait?
I think "to get the idea" you have to be engaged comparing several
traits. What are you comparing them on? What is it supposed to
relate to? - If you want something that is *interval* (better by how
much) rather that *ordinal* (which is better), that is not trivial.
The underlying question is: What is your basis of defining linearity?
The genetic contribution, in at least one sense, is linear in the
squared term. But there is probably another sense where the square
root fits better.
Maybe someone else can help with this other question -
>
> Supplementary question: can someone supply a definition of the
> calculation of heritability in twin studies where you have
> dizygotomous twins separated at birth acting as a kind of control for
> monozygotomous twins separated at birth? The paper I am trying to
> understand is Fox et al. (1996). There is disappointingly
> insufficient detail in the Methods, and they refer to a text book
> that I can't access. I've clicked around the web without success
> looking for an approachable explanation. I'd really appreciate a
> link to a good website on this topic.
>
> Fox PW, Hershberger SL, Bouchard TJ (1996). Genetic and environmental
> contributions to the acquisition of a motor skill. Nature 384, 356-358
>
> Will
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
=================================================================
Instructions for joining and leaving this list and remarks about
the problem of INAPPROPRIATE MESSAGES are available at
http://jse.stat.ncsu.edu/
=================================================================
