Fixing parameters to equality using sam()

[Manuel Rein]

Hello,

I want to fix factor loadings across two factors to equality. However, lavaan seems to ignore that and estimates them freely. Below is an example code:

---------------------------------------------------------------

library(lavaan)
HS.model <- " visual =~ x1 + x2 + x3
textual =~ x4 + l2*x5 + x6
speed =~ x7 + l2*x8 + x9
"
# SEM
sem_fit <- sem(HS.model, data = HolzingerSwineford1939)
summary(sem_fit)

---------------------------------------------------------------

It can be seen that the factor loadings of x5 and x8 are equal (1.125).

However, using the sam() the loadings differ:

---------------------------------------------------------------

# SAM

sam_fit <- sam(HS.model, data = HolzingerSwineford1939, output = "list")
sam_fit$LAMBDA

--------------------------------------------------------------

Here, the factor loadings are 1.133 and 1.225.

Is there a way to fix the loadings to equality?

Kind regards

Manuel Rein

[Jasper Bogaert]

Hi,

I think the default approach in sam() is to use separate measurement blocks for each latent variable. This is not the case when using sem and this could explain the problem. I think it works when you define the measurement blocks (one for visuals and one for textual and speed together) using the mm.list argument. See code below:

sam_fit <- sam(HS.model, data = HolzingerSwineford1939, mm.list = list("visual", c("textual", "speed")))

lavInspect(sam_fit, what = "est")

and see https://rdrr.io/cran/lavaan/man/sam.html

Best wishes, 

Jasper

Op woensdag 2 november 2022 om 12:30:01 UTC+1 schreef Manuel Rein:

[Manuel Rein]

Hi Jasper,

thank you very much, this helped!

Best

Manuel