Error in colMeans when trying to plot model with ordered variables using a self defined matrix layout

[adriaanvl]

Dear Sacha,

Whilst trying to plot a model, I get the following error

Error in colMeans(x, na.rm = TRUE) : 'x' must be numeric

I get this error only in the following combination (both conditions must be present): Condition 1: When the sem(fit, df, ordered=c("X1", "X2"...)) is instructed to treat specific variables as ordered. Condition 2: When in the semPaths command I want to use my own matrix than the standard one provided by the software.

With my data, when I remove either ordered=, or when use one of the predefined layouts (tree, circle, etc.), I do not get the above problem.

I do not wish to burden you with my data or code, so below is an example which reproduces the result (The example, not my data, creates a problem when I don't define specific variables as ordered, but hopefully the below example is helpful).

set.seed(1234)
X1 <- sample(0:2, 100, replace=TRUE)
X2 <- sample(0:2, 100, replace=TRUE)
X3 <- sample(0:2, 100, replace=TRUE)
X4 <- sample(0:2, 100, replace=TRUE)

df <- data.frame(X1, X2, X3, X4)

example.model <- '
latent =~ X1 + X2 + X3 + X4
'

fit <- cfa(example.model,
           df,
           ordered = c("X1","X2","X3","X4")
           )
summary(fit, fit.measures=TRUE, standardized=TRUE)

#create plot
semPaths(fit)

# Create plot using self defined matrix
L <- matrix(
  c(
    NA,       "latent",     NA,     NA, 
    "X1",   "X2",   "X3", "X4") 
  ,4 )

Graph <- semPaths(fit, layout=L)
Graph$graphAttributes$Edges$curve <-
  ifelse(Graph$Edgelist$bidir, 1, 0)
plot(Graph)