# Correlation, Causation & Dr. Barnum

BillJanie@aol.com
Sat, 9 Mar 1996 20:12:45 -0500

--PART.BOUNDARY.0.8468.mail02.mail.aol.com.826420364
Content-ID: <0_8468_826420364@mail02.mail.aol.com.333719>
Content-type: text/plain

Attach file.

--PART.BOUNDARY.0.8468.mail02.mail.aol.com.826420364
Content-ID: <0_8468_826420364@mail02.mail.aol.com.333720>
Content-type: text/plain;
name="TOLOIS3"
Content-Transfer-Encoding: quoted-printable

Lois,
=0D
You pose some interesting questions. The most
crucial, for now, is how do we distinquish causal =

from merely predictive relationships. This really is
the radical departure of corresponding regressions
from path analysis and regression. The distinction
is perhaps best addressed by careful examination
of the notion of a correlated dependent variable.
=0D
In our simulations we model the genesis of =

variables from the addition of variables to form =

larger, more complex structures- or if we want to =

go the direction Rue suggested, more complex
flows of streams of being. Earlier we consider
the distinction between simply clumping variables =

together versus synthesizing them into a higher
order unity. The higher order unity will reveal
operations of actual genesis by which individual
raw sores are generated from other individual
raw scores. This point has been missed by the
path analysis folks, I think because it is so much
easier to just decompose correlation matrices. =

With corresponding regressions, we must have
the raw data, the correlations are not enough .
In procedures like factor analysis and multiple =

regression, the stats can usually be developed =

with nothing but the correlation matrix. True, raw =

scores are needed for factor scores in factor
analysis and for predicted scores in regression =

analysis, but most people using factor analysis
is that if we restrict ourselves to the analysis of =

just correlations, we are dealing with scalars- i.e. =

single indices that represent only certain relations
between the many other numbers (raw scores). =

The raw scores are much closer to the actual =

dynamics of nature. The causal dynamics are
glossed over in the scalar form of correlations.
Corresponding regressions goes deeper into the
raw data. That's how it manages to get at the
causal dynamics that are obscured by mere
correlations.
=0D
In Y=3DX1+X2 we have simulated a formal cause. Y
is the synthesis of X1 and X2. Traditionally folks
have made the assumption that since Y=3DX1+X2,
that X1=3DY-X2. When dealing with scalars this
symmetry is true. But when we look closer at the
actual raw data that is generated- say across 50
rows of simulated observations- we discover patterns
in the raw data that do not support the assumption
that Y=3DX1+X2 and X1=3DY-X2 are equally plausible
explanations of the data. In fact, we did generate =

the 50 observations on the 3 variables using Y=3DX1+X2
=2E We did not use X1=3DY-X2. Y, by definition,
depends on X1 and X2, not the other way around. =

=0D
This point is very relevant to my use of formal
cause. The equation Y=3DX1+X2 has a logical
DEVELOPMENT but it does not depend on time. =

The equation is logically immediate. True when
we use a computer to simulate the data,it takes
time, but this form of efficient causality is
incidental to the essentially timeless nature of
Y=3DX1+X2. It does not take time for Y=3DX1+X2. =

This timelessness does not mean, however, that
Y=3DX1+X2 is the same as X1=3DY-X2. In our =

timeless but real synthesis,Y is defined by its
inclusion of X1 and X2. X1 does not include Y.
Inclusion is the essence of formal cause.
=0D
Correlation does not imply inclusion. The case
of correlated dependent variables demonstrates
this. Recall that earlier we considered a model
in which wisdom and competence had a common =

root in education. This common ground will lead to
a correlation between wisdom and competence,
but neither of these variables causes (is included in)
the other. They do have predictive power regarding
one another. A person who is wise will be educated.
That makes it more likely that he could also be =

competent, since education is necessary for =

competence (in our hypothetical model). But wisdom
does not logically entail competence. The common
ground of education may be there, but other things =

necessary may not be there. Thus, wisdom does not
cause competence, nor competence wisdom, even
though they are correlated. The correlation may
seem important, but really it is potentially just =

deceptive. This is why liars like to use loose
constructs and phrases. They can seem to be =

saying anything because there are so many
correlations between what they say and what =

they want you to think is true. They just leave =

out essential points, points that they may be
unaware of themselves- if they are ignorant or
living in bad faith. Some therapy clients get the
feeling their doctors are doing exactly this. =

=0D
On to your point concerning the linearity of
cause. True, causes can loop back on one
another to create very complex dynamics. =

We should be able to use grids to chase the
whirling causes. We could tack down the =

pattern to particular circumstances. It is likely,
however, that many psychological problems
are immediately linear and this will not be a problem.
I have found actual data from questonnaire
studies of cults that suggest patterns like this:
A causes B, B causes C, and C causes A. =

To use our "cutter" example, A:feeling empty =

causes her to feel B:lifeless. Feeling lifeless is =

added to C: the desire to feel. These combine
with another construct (D) ? makes up the =

E:"cut myself" experience. When the pain of =

the cut recedes, the contrast leads to a
heightened since of A:(Feeling empty). These
kinds of patterns, where there are intervening
variables (B,C,D) between A and E, pose no
problem for corresponding regressions. =

=0D
I touched on the Barnum effect earlier. Such
things may complicate the picture but mere
suggestion is unlikely to make any extensive =

impact on the client.I know there are cases
where it can, for example in relaxation therapy
and desensitization etc. but when dealing with
constructs, merely superficially changing the
clients intepretation of his constructions via a =

kind of slot rattling, is unlikely to amount to much.
Consequently, if you find the true causal
determinant of the clients constructions, be =

prepared to help them discover the fact =

themselves. At least you will know which =

experiments need to be developed, thus saving
much time and confidence. Otherwise, the
client will lose faith, no matter how superficially
impressive the diagnosis and the doctor. =

Sooner or later something clicks, the truth
adds up, and the client who has been duped
gets the picture, if he is lucky. Psychology
then has an enemy.
=0D
Bill
=0D
=

--PART.BOUNDARY.0.8468.mail02.mail.aol.com.826420364--

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%