Dear Stata users,
I am working with an outcome variable asked in a Likert scale, repeated measures for three time points, and all of its distribution is very skewed to 5 with a slight decreasing tendency.



Array

What is the best way to fit a latent growth model with this three-time point outcome variable? Should I use categorical outcome approach (e.g. ordinal, or coding them differently)? or can I just treat it like a normal continuous variable?

When I just try regular latent curve modeling (treating like a normal continuous variable), the fit statistics doesn't seem to work. I am new to the latent growth curve modeling, so anything would help.



. sem (Intercept@1 Slope@0 timevarying1 timevarying2 -> outcome1)(Intercept@1 Slope@1 timevarying1 timevarying2 -> outcome2)(Intercept@1 Slope@2 timevarying1 timevarying2 -> outcome3)(Intercept Slope <- iv1 iv2 iv3 iv4...), cov(e.Intercept*e.Slope) method(mlmv) noconstant

.estat gof, stats(all)

----------------------------------------------------------------------------
Fit statistic | Value Description
---------------------+------------------------------------------------------
Likelihood ratio |
chi2_ms(17) | 33195.020 model vs. saturated
p > chi2 | 0.000
chi2_bs(33) | 10467.027 baseline vs. saturated
p > chi2 | 0.000
---------------------+------------------------------------------------------
Population error |
RMSEA | 0.277 Root mean squared error of approximation
90% CI, lower bound | 0.000
upper bound | .
pclose | 0.000 Probability RMSEA <= 0.05
---------------------+------------------------------------------------------
Information criteria |
AIC | 759595.779 Akaike's information criterion
BIC | 760304.478 Bayesian information criterion
---------------------+------------------------------------------------------
Baseline comparison |
CFI | 0.000 Comparative fit index
TLI | -5.173 Tucker-Lewis index
---------------------+------------------------------------------------------
Size of residuals
CD | 0.469 Coefficient of determination