Standardized residuals and Modification Indices for CFA with ordinal data

[Hellen Geremias dos Santos]

Dear lavaan users,

I am trying to fit a CFA model with ordinal data. 

I have two questions:

First, Is it make sense require standardized residuals in this situation?

I tried to use these commands:

fitcfa1 <-cfa(model1, data=banco1,mimic="Mplus", se="robust",test="scaled.shifted",orthogonal = FALSE, ordered=c("aa1","aa2","aa3","aa4","aa5","aa6","aa7","aa8","aa9","aa10","aa11","aa12","aa13","aa14","aa15","aa16","aa17"),start="Mplus",std.lv=T)

resid (fitcfa1,'standardized')

but I received the following error:  standardized and normalized residuals only availabe if estimator = ML (or MLF, MLR, MLM)

I have doubts on how to interprete unstandardized residual for this analysis. Someone could suggest some literature?

The other question is about modification indices. 

I used the command:

modindices (fitcfa1) 

But the decision column related to these indices weren't shown. I don't understand the results. Someone could suggest some literature here too?

Any help would be much appreciate

Kind regards

Hellen

[Terrence Jorgensen]

Is it make sense require standardized residuals in this situation?

When analyzing ordinal data, the first step is (assuming there is a standard normal variable underlying each ordered variable) to estimate thresholds and the polychoric correlation matrix for the unobserved continuous variables.  Because it is a correlation matrix, the residuals you request are already standardized.  They can't be standardized any further because the variances are assumed to be 1.

Terry

[Hellen Geremias dos Santos]

Dear Terry

Many thanks for your advice on this. 

Kind regards, 

Hellen