Constrain correlations between items to be negative

46 views
Skip to first unread message

Dimitrios Zacharatos

unread,
Feb 5, 2018, 11:28:33 AM2/5/18
to lavaan
I am trying to create a CFA model in lavaan and I do not know how to constrain some parameters

I need for instance the covariance between some observed items to be negative whilst in other cases I need it to be positive

In addition I need to constrain error for items to be the sum of ψ's (e.g. Ψ 2 7 + Ψ 2 8)

As I understand it a CFA can be created in that way

model<-' η2=ψ4 5 + ψ5 6 + ψ7 8 + ψ8 9
               η3=ψ4 6 + ψ7 8 + ψ7 8 + ψ7 8

               η2~~1*η2
               η3~~1*η3

# how I define constraints for negative covariances between items?
# how I define uniqueness so that is equal to sum of 2 utility uniqunesses
'

Screenshot from 2018-02-05 18-18-09.png

Jeremy Miles

unread,
Feb 5, 2018, 11:32:19 AM2/5/18
to lav...@googlegroups.com
This is straightforward to do, but why do you need it to be negative (or positive). The value of the covariance comes from the data. 

Here's how to do it:

y_78 ~~ c1 * y_89 # sets up the covariance and labels it
# constraints
c1 < 0

Note that this also makes your df weird, and so you might not get a chi-square p-value.

Jeremy

--
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.
--
--
My employer has nothing to do with this email. 

Dimitrios Zacharatos

unread,
Feb 5, 2018, 11:37:23 AM2/5/18
to lavaan
You are right covariance comes with the data and that makes no sense. but what does -ψ2 8 means in the plot?

Jeremy Miles

unread,
Feb 5, 2018, 4:06:04 PM2/5/18
to lav...@googlegroups.com

I've never seen that. It implies to me that there's a value psi_28 somewhere, and that this should be constrained to be the negative of it. But there isn't, so I don't know.

Dimitrios Zacharatos

unread,
Feb 6, 2018, 8:26:30 AM2/6/18
to lavaan
Do you think these ψ's could represent correlated errors between items?

Jeremy Miles

unread,
Feb 6, 2018, 12:06:38 PM2/6/18
to lav...@googlegroups.com
I don't know, it's a weird model.
Reply all
Reply to author
Forward
0 new messages