On Tue, 25 Apr 2000 05:32:36 +0100, "haytham siala"
<[EMAIL PROTECTED]> wrote:
> Hi,
> I was just wondering if the factor analysis results show significant
> negative ( > 0.3) factor loadings (correlations) on an item, what am I to
> imply and how do I then calculate the factor score if the same factor loads
> positively on some items and negatively on some others?
A negative loading should not be a problem. In the early days of
scale construction, having negative loadings was something that was
intended and devised, in order to avoid certain artifacts of
test-taking behavior ("response bias" for instance). By counting a
particular set of "Yes" responses and adding to a similar count of
"No" responses, the final count would not be biased by the "tendency
to answer Yes." The same principle was used for items with more than
0/1 scoring, too, by reversing the values to be added -- so a 1-to-5
variable becomes a 5-to-1 variable, when rescored.
In a factor analysis, any old item with a negative loading can have
its score reversed, and then you add it in as usual.
The fact that Negative loadings, on occasion, *seem* to be a
problem, is something that worries me -- it seems to be evidence that
a lot of people really do not understand what is being done (but
usually, we just don't notice).
=============about factoring.
Take note, that a raw, unrotated, factor analysis has as many negative
as positive loadings (approx.) on every factor after the first one.
The factors all *have* to be bi-polar like this, by the manner of
derivation, since each step of stripping out a "factor" or eigenvector
is going to leave the equivalent of a matrix with half the
correlations negative. And the next Factor shows the strengths among
those correlations.
What is the meaning of two positive loadings on one factor? - the two
variables tend to be high (and low) at the same time. And, two
*negative* loadings mean exactly the same thing. The difference, yes,
is with loadings with different signs -- one variable (apparently) is
higher than usual, when the other is lower.
Okay, we usually look at factors after "rotating toward simple
structure." With psychological test items, that usually results in
most Scales having a few large loadings that are all the same sign
(those are arbitrarily written as positive -- by multiplying loadings
by minus one when the program detects more signs in the "wrong"
direction). If you look at the items and you can't figure why the
question is in the "opposite" direction, then maybe you have trouble
with your data file, or maybe you really need a primer on
understanding data.
--
Rich Ulrich, [EMAIL PROTECTED]
http://www.pitt.edu/~wpilib/index.html
===========================================================================
This list is open to everyone. Occasionally, less thoughtful
people send inappropriate messages. Please DO NOT COMPLAIN TO
THE POSTMASTER about these messages because the postmaster has no
way of controlling them, and excessive complaints will result in
termination of the list.
For information about this list, including information about the
problem of inappropriate messages and information about how to
unsubscribe, please see the web page at
http://jse.stat.ncsu.edu/
===========================================================================
