Moderation Analysis SEM with different number of indicators: 2 moderators

[Valeria Rudits]

Dear lavaan community,

I am trying to perform a moderation analysis in R using SEM. However, I am not sure if the approach I have chosen suits my variables, as the latent variables in my model have a different number of indicators. I have 2 IVs: one is observational, and the other has 2 indicators; 2 moderators, one is observational and the other has 3 indicators. I feel a bit lost with the IndProd function due to the different number of indicators and if I should split the model into 2 due to too many interactions. I would appreciate any guidance, thank you. Currently the model I am trying to build looks like that 

A <- paste0(1:2)
B <- paste0(3:5)
C <- paste0(6)
D <- paste0(7)

intNames <- paste0(rep(a,  each = length(B)), B)

dat2 <- indProd(data1_clean, var1 = c(1:2),
                var2 = c(3:5), var3 = (6), var4 = (7)
                match = FALSE, residualC = TRUE,
                namesProd = intNames)

### Residual Centering interaction model

modint <- '
A =~ A1 + A2
B =~ B1 + B2 + B3
DV =~ DV1 + DV2 + DV3 + DV4 + DV5
int1 =~ A1.D + A2.D
int2 =~ C:D
int3 =~ C.B1 + C.B2 + C.B3

#Fix covariances between interaction and predictors to 0
A ~~ 0*int1

D ~~ 0*int1

C ~~ 0*int2

D ~~ 0*int2
B ~~ 0*int3

C ~~ 0*int3

DV ~ A + D + Int1
DV ~ C + D + int2
DV ~ C + B + int3

[Terrence Jorgensen]

I feel a bit lost with the IndProd function due to the different number of indicators and if I should split the model into 2 due to too many interactions

You could model 1 interaction at a time to check which ones are significant and worth probing (or dropping to simplify the model).  Regardless, indProd() only creates product indicators for one 2-way latent interaction at a time.  So you need to call it 3 times: once for each set of product indicators to add to the data set.

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen

[Valeria Rudits]

Thank you for your prompt reply, Dr. Jorgensen. I've tried running the model with an observational and a latent variable (2 factors), however, the model output I receive is missing the standard error and the p-value. What would be the way to could fix this issue? Thank you for your guidance.