path analysis

#1
If I wish to do path analysis, how would I put the these variables in a model?

I have 2 IVs (e.g. A and B), 1 DV (e.g. C). And A,B and C can be measured using questionnaires.
For each variable,
A has 2 sub-constructs, (e.g. subA1 and subA2)
B also has 2 sub-constructs, (e.g. subB1 and subB2)
C can be treated as one variables or separately with 5 sub-constructs (e.g. subC1, subC2, subC3, subC4, subC5)

If I try to do a simple path model, I would use 'subA1, subA2, subB1, subB2' as enxogenous and C as endogenous variables.

Can anyone teach me how would some possible models look like?
 
#2
All models should really have a theoretical under-pinning, you cannot just randomly piece together a model (imo). Are you interested in any of the sub-scales in particular predicting the DV?

When you say a 'simple path model' are you interested in just representing the constructs as observed variables or as latent constructs?

Also, what is the anticipated relationship between your two IV's?
 
#3
Yes. I have a theoretical explanation for the model.
all sub-constructs are observed variables, and A, B, C are latent.

subA1 has -ve corr. to subB1 and subB2
subA2 has +ve corr. to subB1 and subB2
all corr. are significant.

can a model be like....
subA1------>
subA2------>
subB1------> C
subB2------>


(everything pointing to C)
 

hlsmith

Less is more. Stay pure. Stay poor.
#4
I am interested in eventually learning path analysis. To me latent means unmeasured, what does it mean in this context (e.g., exogenous)?
 
#5
there are exogenous and endogenous variables

to explain in a simple way, you may treat exogenous as your IV and endogenous as your DV
 
#6
'can a model be like....
subA1------>
subA2------>
subB1------> C
subB2------>'

That would be a simple regression, or if you are proposing that C is latent, then this also represent a CFA
 
#7
I am interested in eventually learning path analysis. To me latent means unmeasured, what does it mean in this context (e.g., exogenous)?
Exogenous or endogenous variable does not refer to whether the variable is latent or not.
In simple terms, endogenous is typically an outcome variable and exogenous a predictor (as highlighted by heidix). So the value of an endogenous variable is determined by the state of the other variables (exogenous) in the model. The state of an exogenous variable is independent of other variables in the model.

In regards to path analysis, this is an over-arching term for lots of types of analysis.
The OP is actually referring to structural equation modelling, whereby there is a measurement model within the total model. This measurement model determines the latent variables from the observed variables.