covaraince between exogenous variables

[unordered]

Hey @all,

I have several questions:

First of all I want to specify the covariance between exogenous variables

q1: the lavaan-default does not estimate the covariance between exogenous variables?

If lavaan does not estimate the covariance, q2 it is possible to modify the default with:

fixed.x = FALSE

to estimate the covariance.

I have some exogenous categorical variables in my model and it works well. But if i fix my code with

sem(model, data=x, fixed.x = FALSE)

I get an error message:

lavaan ERROR: unordered factor(s) detected; make them numeric or ordered:

I know this error very well from my beginning with lavaan, but I fixed it soon (Tutorial). So the error message only comes, when I modify my code with fixed.x = False

I hope it is not a redundant problem and I am looking forward to your help

thanks,

marco

[Christopher David Desjardins]

On Thu, Nov 15, 2018 at 4:14 AM unordered <team.pro...@gmail.com> wrote:

Hey @all,

I have several questions:

First of all I want to specify the covariance between exogenous variables

q1: the lavaan-default does not estimate the covariance between exogenous variables?

That's correct.

 

If lavaan does not estimate the covariance, q2 it is possible to modify the default with:

fixed.x = FALSE

to estimate the covariance.

You would set fixed.x = TRUE

I have some exogenous categorical variables in my model and it works well. But if i fix my code with

sem(model, data=x, fixed.x = FALSE)

I get an error message:

lavaan ERROR: unordered factor(s) detected; make them numeric or ordered:

You need to fix this and as the message says either make theM factors (look at str(x)) and change them to numeric or ordered variables. This is unrelated to your fixed.x = FALSE questions though.

I know this error very well from my beginning with lavaan, but I fixed it soon (Tutorial). So the error message only comes, when I modify my code with fixed.x = False

I hope it is not a redundant problem and I am looking forward to your help

thanks,

marco

--
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.

[unordered]

Hey Christopher,

it goes forward, but I still have some questions:

I fixed the problem with the categorical variables.

Now I want to be sure about estimation of exogenous variable

lets say i have the follow simple code:

mod <-
'lat_var1 =~ x1 + x2 + x3
lat_var2 =~ x4 + x5 + x6

#Regression
lat_var1 ~ var_1 + var_2
lat_var2 ~ var_1 + var_2

#now i want the covariance between the exogenous variables (var_1 and var_2)
var_1 ~~ var_2

#in the SEM-Function I set fixed.x= TRUE

mod.fit <- sem(mod, data=data, fixed.x=TRUE)

now i estimate freely the covariance between var_1 and var_2 and the results are different from the results without covariance.

There is no difference in the output between

fixed.x = TRUE

#and

fixed.x FALSE

What is the difference in the Argument TRUE and FALSE?

again: thanks a lot for your help!

marco

[Christopher David Desjardins]

It's because you are overriding the fixed.x = TRUE with the exogenous covariance. Presumably, you saw this WARNING message when you specified fixed.x = TRUE.

lavaan WARNING: syntax contains parameters involving exogenous covariates; switching to fixed.x = FALSE

It says that because you are specifying a covariance and it's telling you that specifying a covariance and setting that to TRUE makes no sense and therefore lavaan is overriding your decision. 

Compare:

model <- ' 

  # latent variable definitions

     ind60 =~ x1 + x2 + x3

     dem65 =~ y5 + y6 + y7

  # regressions

    dem65 ~ y1 + y2

    ind60 ~ y1 + y2

  # residual correlations

    y1 ~~ y2

'

fit <- sem(model, data=PoliticalDemocracy)

summary(fit)

to:

model <- ' 

  # latent variable definitions

     ind60 =~ x1 + x2 + x3

     dem65 =~ y5 + y6 + y7

  # regressions

    dem65 ~ y1 + y2

    ind60 ~ y1 + y2

'

fit <- sem(model, data=PoliticalDemocracy, fixed.x = FALSE)

summary(fit)

to:

model <- ' 

  # latent variable definitions

     ind60 =~ x1 + x2 + x3

     dem65 =~ y5 + y6 + y7

  # regressions

    dem65 ~ y1 + y2

    ind60 ~ y1 + y2

'

fit <- sem(model, data=PoliticalDemocracy, fixed.x = TRUE)

summary(fit)