Problems with creating a table for latent factor correlation

[JL]

Hi all,

I have a problem creating a table for latent factor correlation for my SEM with multiple imputation.

I am using a SEM to test the Theory of Planned Behavior. ATT, SN, PBC and INT are likert-scale variables, BEH is ordinary/binary. I use multiple imputation for missing data.

I would like to create a table for latent factor correlations with this code: 

##Creating a table with latent factor correlation

parameterEstimates(modelfitted2, standardized=TRUE) %>% 

  filter(op == "~~", 

         lhs %in% c("ATT", "SNORM", "PBC","INT", "BEH"), 

         !is.na(pvalue)) %>% mutate(stars = ifelse(pvalue < .001, "***", 

                                                   ifelse(pvalue < .01, "**", 

                                                          ifelse(pvalue < .05, "*", "")))) %>% 

  select('Factor 1'=lhs, 

         'Factor 2'=rhs, 

         Correlation=est, 

         sig=stars) %>% kable(digits = 2, format="pandoc", caption="Table 3: Latent Factor Correlations")

(copied from: http://www.understandingdata.net/2017/03/22/cfa-in-lavaan/)

It does work for the normal SEM function:

modelfitted <- sem(model, data = df, ordered = "BEH")

but does not work if I use multiple imputation:

modelfitted2 <- sem.mi(model, data = df, ordered= "BEH", m = 20, seed = 12345)

Then I get the error: 

"Error in ~!is.na(pvalue) : object 'pvalue' not found"

Does anyone have a recommendation how to solve this? Is there another way to get the correlations between the latent factors (instead of using the covariances from the summary table)?

Thanks!

[Terrence Jorgensen]

but does not work if I use multiple imputation:

modelfitted2 <- sem.mi(model, data = df, ordered= "BEH", m = 20, seed = 12345)

That is because parameterEstimates() is not a generic function exported by lavaan, so I could not write a method specific for lavaan.mi objects.  For lavaan objects, the generic summary() method only prints information to the screen from various functions, without being able to save output, so I coopted it by writing a generic summary() method for lavaan.mi objects returns the equivalent of parameterEstimates() output (by default, with the argument add.attributes = TRUE, so that the printed result resembles lavaan's summary() output).

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