On 05/20/2015 12:45 AM,
ac...@umich.edu wrote:
> Were you able to fit an ALT model at all? I keep on having a problem
> where the first time point is estimated by the program instead of
> treated differently from the later time points.
I have fitted several ALT models without any issues. But I recommend to
use the lavaan() function for this, and avoid the growth() function. The
latter is only for simple growth models.
Below is an example from a book chapter written by Curran & Bollen (2001).
Yves.
# example from Curran & Bollen (2001) book chapter `The Best of
# Two Worlds'
# note the negative variance for the slope factor.
# this was easy to miss using EQS, since it forces variances
# to be strictly positive...
lower <- '
2.926
1.390 4.257
1.698 2.781 4.536
1.628 2.437 2.979 5.605
1.240 0.789 0.903 1.278 3.208
0.592 1.890 1.419 1.004 1.706 3.994
0.929 1.278 1.900 1.000 1.567 1.654 3.583
0.659 0.949 1.731 2.420 0.988 1.170 1.146 3.649 '
COV <- getCov(lower, names=c("anti1", "anti2", "anti3", "anti4",
"dep1", "dep2", "dep3", "dep4"))
MEANS <- c(1.750, 1.928, 1.978, 2.322, 2.178, 2.489, 2.294, 2.222)
model <- '
# growth curve
i =~ anti1 + 1*anti2 + 1*anti3 + 1*anti4
i ~ 1; i ~~ i
s= ~ anti1 + 1*anti2 + 2*anti3 + 3*anti4
s ~ 1; s ~~ s
anti1 ~~ anti1
anti2 ~~ anti2
anti3 ~~ anti3
anti4 ~~ anti4
# autoregressive components
anti2 ~ a1*anti1
anti3 ~ a2*anti2
anti4 ~ a3*anti3
'
fit <- lavaan(model, sample.cov = COV, sample.mean = MEANS,
sample.nobs = 180, mimic = "EQS")
summary(fit, fit.measures = TRUE)