On 29 Apr 2002 06:39:03 -0700, [EMAIL PROTECTED] (Andrew) wrote:
>
> My model will predict that the correlation of two exogenous variables
- does this mean,
'My model says that having corresponding values,
both "self" or both "other",
will predict a certain outcome.'
That is the colloquial use of 'correlation' rather
than the statistical one.
Nor would I ever say, statistically speaking,
"cause the *presence* of another *variable*".
A variable is part of a model or it is not;
what you account for is a *value* of the variable,
or you can speak of the presence of a trait.
> ("locus of control" and "agency") will predict or cause the presence
> of another variable. Each of the the two exogenous variables alone
> will not do this: only in the specific case of them being correlated
> will the prediction be in effect.
>
> Would it be possible to place, in the model, an endogenous variable
> consisting of the correlation between the two exogenous variables of
> locus of control and agency (which in turn would affect the other
> variable)? In other words, an endogenous variable consisting of the
> specific way that locus of control and agency interact.
>
If you are asking whether a single variable might represent
(or re-code) an interaction, the answer is Yes.
But I don't know about exogenous and endogenous, since
I don't use those terms.
I could mention, if you are not careful about the measurement,
I expect "locus of control" may have problems like those of
"ego strength" - there is one end that is general *good*, but
the most extreme scores, off that end of the scale, should
belong to raving psychotics. If your sample is that broad.
> I'm going by Rex Klein's book and technically it may fit, but my
> statistical knowledge is admittedly far from perfect.
Hope this helps.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
.
.
=================================================================
Instructions for joining and leaving this list, remarks about the
problem of INAPPROPRIATE MESSAGES, and archives are available at:
. http://jse.stat.ncsu.edu/ .
=================================================================
