education correlation

Dave Long dl@silcom.com
Thu, 30 Jan 2003 12:35:40 -0800


As you all can tell from the subject line, I
am also only talking about correlations, not
causations.  Now, for the purposes of this
thread, I find it perfectly reasonable to say
> People who complete high school are _different_ than those who do not.
but, for the purposes of this thread, I must
ask: "but how different are their children?"

The question is not whether prenatal attitudes
are good predictors of parental attitudes, but
whether one generation's educational attainment
is (should be) the best predictor of the next's. [0]

Let's take a short break and examine two different
extremes of correlation: in one corner, we have
Rosseau and a die (assisted by Fermi, von Neumann,
and Ulam); in the other corner, we have de Maistre
and several lead weights (assisted by Galileo).

Rosseau has the die rolled, and discovers that the
correlation between one roll and the next is so low
that the question "is there a pip in the middle?"
tells one far more about what number is showing
than the question "what number came up last?".

De Maistre has the weights dropped off the top of
a tall building, and discovers that the correlation
between one fall and the next is so great that the
question "how massive is this weight, and how hard
was it pushed (horizontally)" tells one far less
about how long it took to hit the ground than the
question "how long did the last one take?".

Speaking hypothetically, if education were like
dropping weights on the moon, then inexperienced
or uncaring parents would be a condemnation to 
piss poor educational opportunities anyway, and
we'd get the same results (more efficiently?) in
a system where parents bought whatever education
they deemed suitable.

Again speaking hypothetically, if education were
like rolling dice, (and if PotUS were a job noted
for its educational requirements) then those same
inexperienced or uncaring parents could have a
child who grew up to be president. [1]

Looking at the real world, how much correlation
should we expect for epigenetic education?  Say:

  education = f(nature, nurture)
  nurture   = g(parental, scholastic)
  nature    = p(genotype)
  genotype  = r(sire, dam)

and given the noisiness of all of these functions,
as well as regression to the mean (being positive
for the disadvantaged), we might hope that the sins
of the fathers are unlikely to run into the second
generation, let alone the third or fourth.  As it
seems that they do, it implies either very strong
correlations, or some room to improve the leverage
of scholastic nurture.

-Dave

:: :: ::

[0] Plato's Socrates comes to the conclusion that
virtue is not teachable, because, as he points out,
good and wise men (being good parents) could find
the best (or at least a good) teacher to bring up 
their sons to be good and wise, but (anticipating
regression to the mean) dynasties of virtue aren't
frequently found.

Dynastic succession is related to the point at hand:
say I have money, a given phenotype, and a degree,
and I wish all of these to descend to my offspring:
passing the estate to my children is easy enough,
the phenotype may or may not breed true, and they'll
have to earn their own degrees.  (somewhat like how,
in feudal times, a son could inherit a kingdom, but
had to win his own spurs; or how, in ancient times,
kings bought easy horses, but learned geometry the
hard way)

[1] As best I can tell from what Mr. Rogers said, the
local public education system means one can choose one's
parents foolishly and not be completely hosed, as long
as the other parents in the area are competent and care.
(also resulting in the rule: choose a school where one
knows the other parents made the same conscious choice)