Differences in for loop and purrr::map

[cmues]

Hello everyone,

I'm doing some analyses using lavaan. I have a dataset that has a large amount of groups. I wanted to investigate fitting a cfa model to each group individually (yes I'm aware of multigroup CFA).

I wrote this code:

mydata %>%
group_by(country) %>%
nest() %>%
mutate(cfa = map(.x = data, .f = cfa, model = Model_4))

This returns some warnings. Which didnt strike me out of the ordinary. However, when I then tried a for loop:

cfa_list <- list()

for (i in 1:length(levels(data$country))) {
  
  data <- data%>% 
    filter(country == levels(data$country)[i])
  
  cfa_list[[i]] <- cfa(model = Model_4,
                             data  = data)
}

Using the loop, I get no warnings from Lavaan. Is this somehow not equivalent? I tried reproducing it but couldn't. Perhaps with a dataset that also contains a larger amount of groups (10+)?

[mara]

It's really hard to say without knowing what the warnings were. (Ideally you'd be able to reproduce and make a reprex, but the warnings would help at a minimum. )

[cmues]

These are some the 2 warnings:

The warnings in question:

"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."

and

"In lav_model_estimate(lavmodel = lavmodel, lavpartable = lavpartable, ... :
lavaan WARNING: the optimizer warns that a solution has NOT been found!"

I tried replicating the issue. I got some open data from the European Social Survey, but the issue isn't present there. It seems specific to the data I'm working on. It's kind of bizarre!