[R] CFA with lavaan or with SEM

[John Fox]

Dear Daniel,

Oh, I see I forgot to comment on your second specification in my last reply:

> -----Original Message-----
> From: r-help-bounces at r-project.org [mailto:r-help-bounces at r-project.org]
> On Behalf Of David Purves
> Sent: Wednesday, January 23, 2013 10:23 AM
> To: John Fox
> Cc: r-help at R-project.org
> Subject: Re: [R] CFA with lavaan or with SEM
> 

. . .

> 
> model.1 <- specifyEquations()
>  f1 = gam11*a + gam12*b + gam13*c + gam14*d + gam15*e + gam16*g
>  f1 = 1* f1
> 

First, this is backwards: the observed variables depend on the factor, and
not vice-versa; e.g., a = gam11*f1. Second, the factor has an error-variance
parameter; it doesn't depend on itself: V(f1) = 1. As I mentioned in my
previous message, it's easier to use cfa() for this kind of model.

Best,
 John