constraining factors

[susana]

Hello,

Anyone knows whether it is possible to constraint factors in model specification?

Thanks,
Susana

[yrosseel]

On 06/13/2013 07:24 PM, susana wrote:
>
> Hello,
>
> Anyone knows whether it is possible to constraint factors in model
> specification?

What do you mean? Can you provide some context and/or perhaps an example?

Yves.

[Susana Eyheramendy]

yes, below is my model specification. 'f2' is a factor and 'y2' is a 0/1 variable. What I would like to do is to set f2 positive when y2=1 and f2 negative when y2=0. Any suggestions?

model1 <- '#
y1 ~  NA*x
y2 ~ NA*f2
f2 =~ NA*x
#variances and covariances
y1~~f2
y1~~y1
y2~~y2
f2~~1*f2
# intercept
y1 ~ 1
f2 ~ 1
# constrains
y2==1*(f2>0)
'

--
You received this message because you are subscribed to a topic in the Google Groups "lavaan" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lavaan/c0jAjVpJkP4/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lavaan+unsubscribe@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at http://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/groups/opt_out.

--
Susana

[yrosseel]

On 06/13/2013 09:25 PM, Susana Eyheramendy wrote:
>
> yes, below is my model specification. 'f2' is a factor and 'y2' is a 0/1
> variable. What I would like to do is to set f2 positive when y2=1 and f2
> negative when y2=0. Any suggestions?

I have no idea what you are looking for. Factors (like f2) are latent
variables. We assume they are normally distributed, and all we can
'constrain' are the mean and variance of the factor.

Perhaps you could explain what 'f2',y2,x,... stand for, and why you wish
to constrain the factor f2?

Yves.

[Susana Eyheramendy]

y2 is 0/1 variable that I assume distributes as probit, that is why the latent  f2 appears which has a truncated normal distribution that I assume is correlated with y1 which is normally distributed. Does this make sense to you? Do you think it is possible to fit such a model with lavaan?

Many thanks,
Susana

Yves.

--
You received this message because you are subscribed to a topic in the Google Groups "lavaan" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lavaan/c0jAjVpJkP4/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lavaan+unsubscribe@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at http://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/groups/opt_out.

--
Susana

[yrosseel]

On 06/14/2013 03:01 PM, Susana Eyheramendy wrote:
>
> y2 is 0/1 variable that I assume distributes as probit, that is why the
> latent f2 appears which has a truncated normal distribution that I
> assume is correlated with y1 which is normally distributed. Does this
> make sense to you? Do you think it is possible to fit such a model with
> lavaan?

It makes sense, but I'm afraid lavaan can not do this. I would try
general-purpose Bayesian tools (WinBugs or Openbugs or Jags).

Yves.

[Susana Eyheramendy]

Ok, many thanks!

Susana

Yves.

--
You received this message because you are subscribed to a topic in the Google Groups "lavaan" group.
To unsubscribe from this topic, visit https://groups.google.com/d/topic/lavaan/c0jAjVpJkP4/unsubscribe.
To unsubscribe from this group and all its topics, send an email to lavaan+unsubscribe@googlegroups.com.
To post to this group, send email to lav...@googlegroups.com.
Visit this group at http://groups.google.com/group/lavaan.
For more options, visit https://groups.google.com/groups/opt_out.

--
Susana

[yrosseel]

The 'LearnBayes' package contains a function called 'bayes.probit',
which uses a truncated latent variable not unlike your setting. May be a
good starting point.

Yves.