***** To join INSNA, visit http://www.insna.org *****

 

Hi, socnetters.

 

In structural equation modeling (SEM) (I studied SEM by reading "Principles
and Practice of Structural Equation Modeling. 2nd. Rex B. Kline"), there is
'a hypothetical causal network model among interested variables'. And also
there is 'a measurement model that try to measure latent variable'. Surely,
we can test feasibility of hypothesized model (containing both 'hypothetical
network model' and 'hypothetical measurement model') using SEM (specifically
using structural regression model).

 

It seems that following statement is possible using SEM :

 

1. Let's assume there are large trace data of actor's attribute (that
should be changing and measured at many times) and network data (maybe
static) by which actor's attribute spread. Then we can set up structural
equation (like path analysis model) by seeing actor as variable and network
data as hypothetical causal model among nodes. And we can test that network
really reflects actor attribute's dynamics. If not, we should reject that
actor attribute is influenced by solely actor's network.

 

2. In case 1, again let's assume that we can't measure actor attribute
directly. Then we can measure other related measurable variable and then
create actor attribute as latent variable caused by measurable variables.
These directly measurable variables and latent variables can be inserted
into the structural equation model. We can also test this actor attribute
(in fact, factor in confirmatory factor analysis) is really feasible.

 

SEM requires large data, so it's difficult to realize this analysis. But,
SEM and adjacency feasibility test seems too similar to be ignored.

Are there any references related to this idea?

 

Thanks in advance.

 

 

School of Physics and Astronomy, Seoul National University.

Cyram Co., LTD. Algorithm Team.

Se Kwon Kim


_____________________________________________________________________
SOCNET is a service of INSNA, the professional association for social
network researchers (http://www.insna.org). To unsubscribe, send
an email message to [log in to unmask] containing the line
UNSUBSCRIBE SOCNET in the body of the message.