cross lagged panel SEM Error because of high covariances

[Joana Brokelmann]

Dear all,

I am trying to run a SEM. I measured two variables (ABB and PSWQ) at two occasions (t0 and t4) and have one binary covariate (group).

This is how I specified my model so far.

modelPSWQ <- ' 

#measurement model

ABBlog.0 =~ A302.0 + A303.0 + A304.0 + A305.0 + A306.0

ABBlog.4 =~ A302.4 + A303.4 + A304.4 + A305.4 + A306.4

PSWQ.0 =~ A502_01.0 + A502_02.0 + A502_03.0 + A502_04.0

PSWQ.4 =~ A502_01.4 + A502_02.4 + A502_03.4 + A502_04.4

#latent model

ABBlog.4 ~ a*ABBlog.0

PSWQ.4 ~ p*PSWQ.0

ABBlog.0 ~ b*PSWQ.0

ABBlog.4 ~ c*PSWQ.4

ABBlog.4 ~ g*group

PSWQ.4 ~ h*group

#indirect effect

hc := h*c

#total effect EAP

total := g + (h*c)

# variances and covariances 

ABBlog.0 ~~ ABBlog.0 

ABBlog.4 ~~ ABBlog.4 

PSWQ.0 ~~ PSWQ.0

PSWQ.4 ~~ PSWQ.4

ABBlog.0 ~~ ABBlog.4

PSWQ.0 ~~ PSWQ.4'

However, if I test the model:

fit <- sem(modelPSWQ, data=ds_wide)

I get the error:

"Error in lav_samplestats_icov(COV = cov[[g]], ridge = ridge, x.idx = x.idx[[g]],  : 

 lavaan ERROR: sample covariance matrix is not positive-definite"

From my research so far I figured that this error probably occurs because some of my variables/items are covarying too high. 

Still, it is logical that the same item, measured at two measurement points, has high covariance.

Do I get the error right? If so, how can I change the specifications of my model without deleting items/variables, as the high covariance is not an error in the data (I checked the covariance/correlation matrix of my model and there are no implausible values)?

Or is there an error in my model/code?

Thank you very much for your help!