lavaan WARNING: the optimizer (NLMINB) claimed the model converged, but not all elements of the gradient are (near) zero; with WLSMV

[Jan Szczypiński]

I try to fit a model of BIS-11 using WLSMV estimator.

The model:

model = '

attentional =~ bis5+bis9+bis11+bis20+bis28+bis6+bis24+bis26
motor =~ bis2+bis3+bis4+bis17+bis19+bis22+bis25+bis16+bis21+bis23+bis30
nonplaning =~ bis1+bis7+bis8+bis12+bis13+bis14+bis10+bis15+bis18+bis27+bis29

'

 fit = cfa(model, data = BIS_data, estimator = "WLSMV")

The fit function results in a warning:

Warning message:
In lav_model_estimate(lavmodel = lavmodel, lavpartable = lavpartable,  :
  lavaan WARNING: the optimizer (NLMINB) claimed the model converged,
                  but not all elements of the gradient are (near) zero;
                  the optimizer may not have found a local solution
                  use check.gradient = FALSE to skip this check.

When I use the summary function i get another warning 

Warning message:
In .local(object, ...) :
  lavaan WARNING: fit measures not available if model did not converge

and the information that

lavaan 0.6-5 did NOT end normally after 129 iterations"

It doesn't converge with any estimator based on DWLS, however it converges with ML estimators.

What's the reason for this? It's important for me to use the WLSMV.

Thanks in advance,

Jan Szczypiński

[Terrence Jorgensen]

It doesn't converge with any estimator based on DWLS, however it converges with ML estimators.

What's the reason for this? It's important for me to use the WLSMV.

That's a lot of categorical items.  How many categories?  If you have a small N and many thresholds, then there might just be too many parameters for your data, whereas MLE would have more stable estimation (single step) and fewer parameters (1 intercept instead of k-1 thresholds per item).

Terrence D. Jorgensen

Assistant Professor, Methods and Statistics

Research Institute for Child Development and Education, the University of Amsterdam

http://www.uva.nl/profile/t.d.jorgensen

[Jan Szczypiński]

Sorry for such a long response. I was gone on vacations for the last 3 weeks. In turned out that it was caused by some missing values in the data structure.

Now, when it's fixed, everything goes smoot.