Random-intercept cross-lagged panel model with categorical and continuous variables

[Cisem]

I am working on a random-intercept cross-lagged panel model to examine the reciprocal relationship between two variables. One of these variables is categorical (likert-scale) and the other one is continuous. Each variable is assessed 5 times. The missingness is MCAR and data is non-normally distributed. I have no observations for some levels of categorical variables (e.g., frequency of “very frequent” is 0, indicating no person crossed this option). My dataset is called “participating.” I used the John's tutorial (https://jflournoy.github.io/2017/10/20/riclpm-lavaan-demo/) while preparing the model (shown in the figure) and informed lavaan about the categorical variables, changed the estimator type and missingness with WLSMV and pairwise, respectively:

 

RICLPM<- lavaan(RICLPModel1, participating,

     ordered = c("soc.1", "soc.2", "soc.3", "soc.4", "soc.5"),

                  estimator = “WLSMV”,

                  missing = “pairwise”,

                  int.ov.free = F,

                  int.lv.free = F,

                  auto.fix.first = F,

                  auto.fix.single = F,

                  auto.cov.lv.x = F,

                  auto.cov.y = F,

                  auto.var = F)

summary(RICLPM, standardized = T)

When I run the script, I got this error message:

 

lavaan WARNING: the optimizer warns that a solution has NOT been found!lavaan 0.6-2 did NOT end normally after 3262 iterations

** WARNING ** Estimates below are most likely unreliable

 

  Optimization method                           NLMINB

  Number of free parameters                         44

  Number of equality constraints                     2

 

  Number of observations                           317

  Number of missing patterns                        37

 

  Estimator                                       DWLS

  Model Fit Test Statistic                          NA

  Degrees of freedom                                NA

  P-value                                           NA

NaNs producedNaNs producedNaNs produced

Parameter Estimates:

 

  Information                                 Expected

  Information saturated (h1) model        Unstructured

  Standard Errors                           Robust.sem

 

Could you help me detect what is wrong with this script? I tried to free intercepts by int.ov.free = T; but I got the same error message. It seems that I am missing something.

Many thanks!