MIMIC model

[kih...@gmail.com]

Dear all,

I want to estimate a model MIMIC. this is a special case of SEM models. unlike the exepmle that I saw on the "lavaan" package, the MIMIC model puts the observed variables in relation to a single latent variable. I have attached an image of the structure of MIMIC. How can I write the program to estimate my MIMIC?

Thanks for your help

YEO

[Jeremy Miles]

What example did you see on the lavaan package? 

 

The model for the diagram you posted is something like:

eta =~ y1 + y2 + y3

eta ~ x1 + x1 + x3

--
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 post to this group, send email to lav...@googlegroups.com.
Visit this group at https://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/d/optout.

[kih...@gmail.com]

the example in this document (page 18)

[Jeremy Miles]

On Mon, 6 Aug 2018 at 03:17, <kih...@gmail.com> wrote:

the example in this document (page 18)

I don't see a mimic model on page 18. Did my previous answer help?

[Kolotioloma YEO]

there is no MIMIC example in the document.

on page 18 this is an example of SEM with lavaan. Since MIMIC is a variant of the SEM models I was trying to understand how it works under Lavaan.

 

De : Jeremy Miles
Envoyé le :lundi 6 août 2018 14:18
À : lavaan
Objet :Re: MIMIC model

 

 

On Mon, 6 Aug 2018 at 03:17, <kih...@gmail.com> wrote:

the example in this document (page 18)

 

 

 

I don't see a mimic model on page 18. Did my previous answer help?

--

[Kolotioloma YEO]

in the framework of MIMIC how to specify residual covariances in the code?

 

 

De : kih...@gmail.com
Envoyé le :lundi 6 août 2018 10:17

À : lavaan
Objet :Re: MIMIC model

 

the example in this document (page 18)

[Terrence Jorgensen]

in the framework of MIMIC how to specify residual covariances in the code?

There is only one operator to specify any (co)variances, residual or exogenous:  ~~

http://lavaan.ugent.be/tutorial/syntax1.html

Since MIMIC is a variant of the SEM models I was trying to understand how it works under Lavaan

You just regress the factor(s) on the exogenous observed predictors using the ~ operator.  Using the ?cfa help-page example, you can add sex and age as predictors of the three constructs:

HS.model <- ' visual  =~ x1 + x2 + x3
              textual =~ x4 + x5 + x6
              speed   =~ x7 + x8 + x9
visual + textual + speed ~ ageyr + sex
'

fit <- cfa(HS.model, data=HolzingerSwineford1939)
summary(fit, fit.measures=TRUE)

Internally, lavaan uses the LISREL representation, so there is no separate vector for the effects of exogenous predictors like there is in the Mplus model:

http://www.statmodel.com/download/muthen1.pdf

Instead, lavaan essentially treats the exogenous predictors of latent variables as though they are single (error-free) indicators of latent constructs, and it automates that process so users do not have to specify exogenous predictors as single-indicator constructs themselves.  The summary() output does not display this explicitly, but you can see it when you view the estimates in the parameter matrices:

lavInspect(fit, "est")

Notice that the columns of $lambda have latent variables corresponding to the exogenous predictors (age and sex) with loadings == 1, and their residual variances in $theta are fixed to zero (assuming no measurement error).  The (co)variances of exogenous variables are in $psi with the other latent variables, and those will simply be the observed sample statistics when fixed.x = TRUE.  This is statistically equivalent to the Mplus model, which has a dedicated parameter matrix (Kappa) for estimating effects of exogenous predictors on endogenous latent factors. 

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

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

UvA web page: http://www.uva.nl/profile/t.d.jorgensen

[Kolotioloma YEO]

what to do when we have negative variances?

thanks in advance

De : Terrence Jorgensen
Envoyé le :mercredi 8 août 2018 10:43

À : lavaan
Objet :Re: MIMIC model

 

in the framework of MIMIC how to specify residual covariances in the code?

--

[Terrence Jorgensen]

what to do when we have negative variances?

Does this have anything to do with MIMIC models having negative variance estimates?  Several other threads have discussed this issue, e.g.:

https://groups.google.com/d/msg/lavaan/ESg_N66yYw8/p84RbJApCgAJ

Terrence D. Jorgensen

Postdoctoral Researcher, Methods and Statistics

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

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