[ExI] Impressive book: Farewell to Alms

hkhenson hkhenson at rogers.com
Thu Feb 7 08:02:51 UTC 2008

At 09:53 PM 2/6/2008, samantha wrote:

>On Feb 4, 2008, at 8:22 PM, hkhenson wrote:
> > At 12:16 PM 2/4/2008, BillK wrote:
> >> On Feb 4, 2008 4:54 PM, hkhenson wrote:
> >
> > snip
> >
> >> There just aren't enough wills compared with the population numbers.
> >> People who wrote wills are a small, self-selected subset of the
> >> population. It is a great leap to conclude that the small subset of
> >> wills is representative of the whole population. The vast majority of
> >> the population had lives that never got into the records anywhere.
> >
> > Elemental sampling theory treats numbers as large
> > as the population as effectively infinite.  The
> > statistical error you get out of samples is such
> > that a sample of a few hundred gives decent
> > expected error bounds on the whole
> > population.  If it's a leap of faith the Gallop polls make the same
> > leap.
> >
> > Do you agree with Clark's numbers for the will
> > data he had?  If you do, what reason do you have
> > to expect the results *not* to apply to the population as a whole?
>Elementary sampling theory requires that the sample is random over the
>population.  The objection was that those who wrote wills in those
>times were not randomly representative of the population as a whole.

I already went over this with Damien here:


>A random sample of wills may say a fair amount about the subset of the
>population who had a need and interest in writing a will while saying
>very little conclusive about the population as a whole.

Clark wasn't trying to sample the population as a whole.  He didn't 
for example give a percentage of the population by asset class.  What 
he was looking for was differential reproduction as a function of 
assets.  To do that he had to analyze wills for assets, children and 
literacy.  Long as he had enough examples for each class to be 
statistically reliable that generated a plot point.  Look at table 4 
in the research report.  From the round number at the end of the 
table I suspect he had lots of wills for the poorest economic class 
that he didn't analyze and use.  But looking at a hundred wills was 
enough to give him a data point.

>Getting a
>good random sample of the population of interest is not a leap of
>faith.  Some care needs to be exercised and a reasonable case made for
>having satisfied this requirement before any statistical argument
>based on such a sample has much merit.

For what he was trying to do, namely show that the rich left more 
surviving children, I think he made the case.  In an era where there 
was no birth control inside marriage, having the assets to feed and 
clothe the children would make a big difference in their survival.

As to extending this result to the rest of England, the observations 
make sense.  The humans were similar in other parts and so was the 
environment.  Can you come up with a reasonable argument as to what 
would make this region different from others?


More information about the extropy-chat mailing list