Moderated Correlation Between Two Second-Order Latent Variables

70 views
Skip to first unread message

Evan Warfel

unread,
Jan 31, 2024, 10:41:41 PM1/31/24
to lavaan
Hi, 

I have a (bunch of) second-order bivariate latent growth curve models for some dyadic analyses I'm running. I'm interested in the correlation between the latent slopes, and in particular, whether this correlation is moderated by the age of the age of the second dyad member. 

In the original model, I have code like the following:
"
<...whole bunch of lavaan code...>

Dyad1LatentIntercept =~ 1*Wave1_Dyad1 + 1*Wave2_Dyad1 + 1*Wave3_Dyad1
Dyad2LatentIntercept =~ 1*Wave1_Dyad2 + 1*Wave2_Dyad2 + 1*Wave3_Dyad2

Dyad1LatentISlope =~ 1*Wave1_Dyad1 + 2*Wave2_Dyad1 + 3*Wave3_Dyad1
Dyad2LatentISlope =~ 1*Wave1_Dyad2 + 2*Wave2_Dyad2 + 3*Wave3_Dyad2

Dyad1LatentIntercept ~~ Dyad1LatentIntercept
Dyad1LatentSlope ~~ Dyad2LatentSlope"

I then fit this model with the growth() function. Can anyone point the way to the best way about how to see if `Dyad1LatentSlope ~~ Dyad2LatentSlope` is moderated by `Dyad2AgeAtWave2`?

The following is not correct, but might be in the right direction?

"Dyad1LatentSlope ~~ Dyad2AgeAtWave2 := Dyad2LatentSlope"

Note that Dyad2Age is a manifest variable. I have looked at inprod, but I don't see how it would help with these second-order latent variables. Any and all help is very much appreciated!

Evan Warfel

unread,
Feb 2, 2024, 1:57:08 AM2/2/24
to lavaan
I have found the article Testing and Interpreting Latent Variable Interactions Using the semTools Package (https://www.mdpi.com/2624-8611/3/3/24), about using the product-indicator approach. It seems very promising. I suppose my question is whether this will work for second-order latent variables?

Many thanks in advance for your attention to my questions!



Reply all
Reply to author
Forward
0 new messages