Modeling mean structure / identification issue

[Thomas]

Hi all,

Relatively new to lavaan. I am trying to conduct a latent change score analysis with two time points for my latent variable (depression) and two indicators per time point. Below is my code. From fiddling I have found if I delete the code specifying the mean structure, everything runs just fine. Even if I just delete the last line, or constrain the intercept of dm2 to be zero, it runs fine. However, as is I get a positive definite error. By my count my model is not under-identified, but am I wrong? And if so, can you help my understand my mistake? Or, is it a syntax error? Any help is appreciated!

lcs_dep <- '
  # measurement model
    dep1 =~ hamd1 + phq91
    dep2 =~ hamd2 + phq92
    dep1 ~~ 1*dep1
    dep2 ~~ 0*dep2
    hamd1 ~~ sigma1_u*hamd1
    hamd2 ~~ sigma1_u*hamd2
    phq91 ~~ sigma2_u*phq91
    phq92 ~~ sigma2_u*phq92
   
  # structural model
    dep2 ~ 1*dep1
    dm2 =~ 1*dep2
    dm2 ~~ dep1
    dm2 ~~ dm2

  # Mean Structure
    dep1 + dep2 ~ 0*1
    hamd1 + hamd2 + phq91 + phq92 + hama1 ~ 0*1
    dm2 ~ 1
'