Re: Underestimating 'r'
- From: "Jim McGinn" <jimmcginn@xxxxxxxxx>
- Date: Wed, 5 Oct 2005 01:00:35 -0400 (EDT)
Perplexed in Peoria wrote:
> "Tim Tyler" <tim@xxxxxxxxxxx> wrote
> > Such truncated trees tend to give lower values for relatedness
> > than using a full tree would give.
>
> True. But do they give a worse value?
This is a really dumb question since you have no basis for judging.
Do they give a value which
> is less usefully employed in Hamilton's rule?
You say useful. But what you really mean is 'appears to explain the
origins of altruism'. IOW, you think that if it appears to explain the
origins of altruism then it is, by definition, more useful. This is an
amateurish and plainly nonscientific approach. You have the cart
before the horse. Accuracy is the only thing scientists should be
concerned with.
>
> If we are looking at, for example, social behaviors which take place
> between siblings - two cheetah cubs from the same litter, say, there is
> no point in not truncating the tree. The right r to use is 1/2, even
> if the population is somewhat inbred. The cub is competing within that
> inbred population. It will have no particular good will toward a
> third cousin if every other cheetah is a third cousin. And its good
> will toward its sibling is not increased because its sibling is also
> another third cousin.
Perfect nonsense.
>
> > How much lower is a question difficult question. If calculating
> > relatedness using hamming distnaces, the figure depends to some extent on
> > the size of the units being compared - and how close a match is needed
> > before two units are described as being related.
> >
> > Also, the figures may be significantly different for groups of organisms
> > like cheetahs (an inbred group) and mice (an outbred one).
> >
> > If "r" between organisms in a group or species is *actually* higher
> > than is commonly belived, that might contribute to group/species-level
> > selection - the extent of the power of which is still the subject
> > of some controversy and debate.
>
> You seem to be agreeing with McGinn here that the absolute degree of
> similarity between donor and recipient is important in the justification
> for altruism. It is not.
Empirically it is the only genetic measure that is relevant to the
issues that Hamilton approached.
What is important is the degree to which
> the recipient is more closely related to the donor than is a member
> of the general breeding and competing population.
Possibly true. But how do we calculate it. And why in the world would
you assume that IBD has anything to do with it. (Not that I'm under
the slightest illusion that PoP will actually answer any of these
questions.)
>
> > Hamilton's rule talks about the circumstances under which a trait
> > will spread through a population - but it doesn't itself consider
> > the possibility of populations competing with one another - and the
> > possibilty of high level selection trumping the effects of low-level
> > selection.
> >
> > So - is "r" higher than convention would dictate; and if so - how
> > much higher? Evidence I've previously posted relating to the
> > frequency of SNPs (and other polymorphisms) suggests to me that
> > it is higher - and significantly so - perhaps enough to make me
> > 70% related to my mother - rather than the 50% convention
> > dictates - and enough to make individuals quite a bit more
> > than 0% related to other "unrelated" members of the population.
>
> Well, if you use the 1970, regression, version of r, your relatedness
> to a randomly selected member of the population is zero BY DEFINITION.
> The fact that the untruncated IBD calculation of r gives a value
> slightly greater than zero indicates a deficiency in the IBD
> definition of r.
How do you discover a deficiency in something the rationale for which
you are unable to explain?
> It does not indicate that altruism should be
> more common than we might think.
How do you supposedly know this? Do you have a crystal ball?
> There can not be any kin-selection
> justification for even the slightest amount of altruism to randomly
> selected conspecifics, regardless of how inbred the species is.
And we're all supposed to take your word for all this?
>
> > If so, what's the practical effect of this? As far as I can
> > tell, the effect would be pretty limited. It ought to make
> > individuals behave somewhat altruistically to other members
> > of their species. However there are a raft of other theories
> > that predict this sort of thing (e.g. reciprocal altruism) -
> > so the effect may be hard to tease out. It wouldn't make
> > *much* difference to how relatives are treated - since
> > their relative relatednesses would remain pretty much
> > the same.
> >
> > Any comments about all this?
>
> Google in this group for r_ibd_hat to see some of my previous postings
> on this subject.
>
> Also see discussion of this topic between Guy and myself on the thread
> "Kin Selection Contradiction" around June 24, 2004.
>
> > What's your personal estimate
> > of "r" between, say, randomly-selected humans?
>
> Regression (1970) r = 0.0000000...
> IBD (1964) r = approximately 2 /6,446,131,399 as of July 2005
> http://www.cia.gov/cia/publications/factbook/rankorder/2119rank.html
>
> Hmmm. Actually these figures are for random selection 'with replacement'.
> You probably intended 'without replacement'. So decrease both estimates
> by 1/6,446,131,400 or so
Pure nonsense.
Jim
.
- References:
- Underestimating 'r'
- From: Tim Tyler
- Underestimating 'r'
- Prev by Date: Re: Underestimating 'r'
- Next by Date: Re: NS and AaD curves
- Previous by thread: Re: Underestimating 'r'
- Next by thread: Re: Underestimating 'r'
- Index(es):
Relevant Pages
|
