HELP! SEM and path analysis with time frames

P

pickle77

New Member
#1
#1
I am having trouble writing the methods section for my dissertation research. I am using SEM/path analysis, and I cannot figure out how to include multilevel analysis for one of my IVs. One of my IVs has 6 levels of time associated with it. The outcome on the DV is being compared between each of these 6 time frames. How do I account for this in the model? Thanks for any suggestions that you can provide.
 
Last edited:
Lazar

Lazar

Phineas Packard
#2
#2
Can you tell us a) whether you are using manifest or latent variables b) what program you are using to estimate the model.
 
Lazar

Lazar

Phineas Packard
#4
#4
There are two ways around this. First, you could dummy code time and put those variables into your model and maybe add interactions between those dummies and any variables whose relationship with the outcome you think might vary by time. This is pretty messy and is not really ideal for a number of reasons (see Gelman and Hill)! Second you could estimate a multilevel model (also known as mixture or HLM) in which Time is a level 1 variable and all other variables are level 2 variables. You could then have varying slopes for the intercepts (that is the intercept is different for each time wave) and maybe varying slopes (the effect of x on y is different for each time wave).

My understanding EQS has no mixture modelling ability but I believe Open MX does have some capacity in this area.
 
spunky

spunky

Can't make spagetti
#5
#5
Lazar said:
My understanding EQS has no mixture modelling ability but I believe Open MX does have some capacity in this area
Click to expand...
both can do them, although for EQS you need the newest version for it. if you dont have it, you can sort of "trick" the older versions thorugh using Raykov's technique of phantom factors... but it is sort of a pain in the !@$##! and should not be attempted unless you have a good knowledge of EQS' language, SEM and an afternoon to waste because you need to do the same process a few times before you get it right.
 
You must log in or register to reply here.