Why R-error "NAs are not allowed in subscripted assignments" when running the modindices command to inspect improvements of SEM?

Question

I tested the measurement model as the first step of my structural equation approach using the R-lavaan package.

Everything works out fine, but when I try to inspect the modindices of that measurement model I get an error message that I do not seem to be able to solve:

    rosiesTAM_measurement<- '

    #measurement model
    PEoU =~ 1*TAM_PEoU_1 + TAM_PEoU_2 + TAM_PEoU_3 + TAM_PEoU_4
    PU =~ 1*TAM_PU_1 + TAM_PU_2 + TAM_PU_3 + TAM_PU_4
    E =~ 1*TAM_E_1 + TAM_E_2 + TAM_E_3 + TAM_E_4
    SI =~ 1*TAM_SI_1 + TAM_SI_2 + TAM_SI_3
    #residual variances
    TAM_SS ~~ TAM_SS
    PEoU ~~ PEoU
    PU ~~ PU
    E ~~ E
    SI ~~ SI
    TAM_SS ~~ PEoU
    TAM_SS ~~ PU
    TAM_SS ~~ E
    TAM_SS ~~ SI
    PEoU ~~ PU
    PEoU ~~ E
    PEoU ~~ SI
    PU ~~ E
    PU ~~ SI
    E ~~ SI
  '
   
   #fit the model
   rosiesTAM_measurement_fit <- cfa(rosiesTAM_measurement_fam_class4, data = rosie_fscores) 
   
   #print summary
   summary(rosiesTAM_measurement_fit, standardized = T, fit.measures = T)
   
   #visualize measurement model
   semPaths(rosiesTAM_measurement_fit)
   
   #check modindices
   modindices(rosiesTAM_measurement_fit, sort = TRUE)
    Error in tmp[cbind(REP$row[idx], REP$col[idx])] <- lavpartable$free[idx] : 
    NAs are not allowed in subscripted assignments
   

The variables relevant for the SEM do not entail any missing values. I do not understand why R does not want to run the modindices command. The measurement model fit is close to be acceptable (sample size was realtively low with N = 183):

• Chi-Square: 0
• CFI: 0.938
• RMSEA: 0.088
• SRMR: 0.084

I appreciate your help, thanks in advance.

Answer 1

The problem was solved! The issue laid in the manifest (observed) variable TAM_SS. When removing this variable from the measurement model, modification indices could be inspected.