2 options, which one is the good one ?

[Sacha Varin]

Dear lavaan R experts,

Good afternoon, 

I write to you for 1 question and a R request.

I have 7 variables. One of them (X_7) is simply the mean of the 6 others (X_1 ; X_2 ; X_3 ; X_4 ; X_5 ; X_6). There is determinism in my data.

Fitting à SEM model, I hesitate between these 2 options.

Either 

X_7 is a latent variable, which is estimated, not calculated as the mean of the others. To be able to assess a mean for this variable, one of the (unstandardized) loadings of the other variables on X_7 is fixed to 1 (per default it is the first). The other loadings are then estimated and the latent variable gets the same unit as the variable that has the fixed loading. So fitting a SEM on this data with X_7 as latent variable.

Or I would use the effects-coded method of scaling (Little, TD, Slegers, DW & Card NA, (2006). A non-arbitrary method of identifying and scaling latent variables on SEM and MACS models. Structural equation  modeling, 13(1), 59-72.)

X1 = 6 - x2 - x3 - x4 - x5 -x6; The mean of latent factor would be the mean of the items ?

Which option is the good One and How to code my SEM model using lavaan R package ?

Best,

Sacha VARIN 

[Terrence Jorgensen]

One of them (X_7) is simply the mean of the 6 others (X_1 ; X_2 ; X_3 ; X_4 ; X_5 ; X_6). There is determinism in my data.

Why do you want a model that has both?  What inference/hypothesis are you trying to draw/test?

X1 = 6 - x2 - x3 - x4 - x5 -x6; The mean of latent factor would be the mean of the items ?

If you have complete data, then the factor mean would be the same as the mean of a scale composite calculated by averaging the items.

Which option is the good One

Again, good to do what?

How to code my SEM model using lavaan R package ?

See the effect.coding= argument in the ?lavOptions help page for automated effects coding.

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

[Sacha Varin]

Dear Terrence,

Really appreciate your answer. I have to explain why I want to fit a SEM model on my data.

I have learned a DAG and I am trying now to estimate effects. In my opinion I can use a SEM to estimate a DAG but it requires me to correctly specify  the functional form of each relationship.

[Terrence Jorgensen]

In my opinion I can use a SEM to estimate a DAG 

SEM can be used for some DAGs, but not if the variables are collinear.

[Sacha Varin]

Dear Terrence,

Yes, thanks for your response. Multicollinearity is a data problem. I use VIF and all VIFs are lower than 3, so collinearity should not be an issue.

Best 

[Paul Joseph]

I greet you well, my question might be trivia. Can I use time series data for SEM especially path analysis. If yes, do I need to text for stationarity? Really need your help

Show quoted text

--
You received this message because you are subscribed to the Google Groups "lavaan" group.
To unsubscribe from this group and stop receiving emails from it, send an email to lavaan+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/lavaan/64907fa3-0250-4e4c-bf38-8dad9cac8ac4n%40googlegroups.com.

[Sacha Varin]

Dear lavaan experts,

In attached file the DAG I get. Some edges (-  -  -  ->) are functional dependencies and the other (---->) are probabilistic dependencies. 

"Score final ISQ 2018" is a latent variable, which is estimated, not calculated as the mean of the others

I have never used lavaan. I was wondering if the R code here below was good according to my DAG as I am trying to estimate effects ?

#####################
library(lavaan)

model <- '

  # measurement model

Engag des élèves  =~ score final ISQ 2018
Engag des enseignants  =~ score final ISQ 2018
Engagement des parents =~ score final ISQ 2018
Efficience =~ score final ISQ 2018
Efficacité  =~ score final ISQ 2018
Équité =~ score final ISQ 2018

  # regressions
    
Efficacité ~ équité
Efficacité  ~ engagement des parents 
Efficience ~ efficacité 

'

fit <-sem(model, data=data)

summary(fit, standardized = TRUE)

##################