Significance levels of in/direct effects in AMOS (with missing data)

#1
Hi,
Is there an alternative was another way to find significance levels, confidence intervals, etc. with missing data information in AMOS (I'm using the newest version), since it does not provide this info in the standard output/estimates when there is missing data? And if not, could someone please recommend an alternative method to getting the significance levels?

FYI - 1) I do NOT have access to Mplus or any other SEM software, and 2) I've run a Sobrel test for mediation significance, but I really need the significance of the other paths as well.

THANKS in advance for any help with this!
 

Lazar

Phineas Packard
#3
You have access to lavaan as it is a free package in R (this will deal with missing data using FIML and provide sig tests etc.). That would work fine. I am not sure why you do not get this info from AMOS however. Does it not just use FIML to deal with missing data?
 

spunky

Can't make spagetti
#4
You have access to lavaan as it is a free package in R (this will deal with missing data using FIML and provide sig tests etc.). That would work fine.
i concur. lavaan is the way to overcome the inherent limitations that AMOS has. if you have internet access, you have access to lavaan
 
#5
THANK-YOU SO MUCH to both of you! I didn't think about free access software. On my way to find and use Lavaan now!

You have access to lavaan as it is a free package in R (this will deal with missing data using FIML and provide sig tests etc.). That would work fine. I am not sure why you do not get this info from AMOS however. Does it not just use FIML to deal with missing data?
Yes, AMOS does use FIML, but while it calculates most other estimates, it specifically states that it does not provide significance or confidence intervals for direct, indirect paths when there is missing data. Odd, but true.
 

Lazar

Phineas Packard
#6
weird.

p.s. If you are not using bootstrapping (which you should) you can just calculate the uncertainty estimates for yourself using the sobel test.
 
#7
Yeah - AMOS also doesn't do bootstrapping with missing data. -_-

I used the Sobel test for indirect effects. I need another way to ensure accurate calculations of the significance of total effects.
 

Lazar

Phineas Packard
#8
well the total effect is just a model in which you have x -> y right (i.e. a model excluding the mediator/s m)? So you just run that model separately.