<html>
<head>
<meta content="text/html; charset=UTF-8" http-equiv="Content-Type">
</head>
<body bgcolor="#FFFFFF" text="#000000">
<div class="moz-cite-prefix">On 2015-09-11 01:09, Rafal Smigrodzki
wrote:<br>
</div>
<blockquote
cite="mid:CAAc1gFhQ57gD5z3ndVfGwgYRnOGjvon=hdv369suJtEErqqK7A@mail.gmail.com"
type="cite">
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
<div dir="ltr"><br>
<div class="gmail_extra">
<div class="gmail_quote">
<div>Are you sure the variance would stay the same? Wouldn't
removing <span
style="color:rgb(51,51,51);font-family:Arial,sans-serif;text-align:-webkit-right">99.9570883466%
of the population from the reproductive pool lower
diversity?</span></div>
</div>
</div>
</div>
</blockquote>
<br>
Yes, but my crude simulation did not measure the variance decline.
Obviously if we select just the two best individuals and have their
offspring be the population diversity is going to be very small.<br>
<br>
The selection formula states that R = h^2 S, where S is the
difference in the mean trait in the selected parent generation to
the population, h^2 is heritability and R the shift in new
population mean. So for IQ, if S=150-100=50 and h^2=0.5-0.8 we get R
as 25-40.<br>
<br>
The variance change depends on the way the genes affect the trait,
so it is not entirely obvious. If the population has no inbreeding,
then removing X% of individuals will remove X% of all genes. If each
gene is has a N(0,1) effect on the trait and there are M genes
affecting the trait, we should expect the trait to be distributed as
N(0,M). Removing X% doesn't change this, but the genes of different
individuals now become correlated. <br>
<br>
<br>
<pre class="moz-signature" cols="72">--
Dr Anders Sandberg
Future of Humanity Institute
Oxford Martin School
Oxford University
</pre>
</body>
</html>