lavaan update breaks glance.lavaan

[yrosseel]

The glance.lavaan() function (in R/lavaan-tidiers.R) is accessing internal slots of a fitted lavaan object to extract some information, instead of using the extractor functions. As the internal slots will change in the next update, this breaks the glance.lavaan() function.

The urgent issue is line 136

converged = x@Fit@converged

which should be replaced by

converged = lavInspect(x, "converged")

as the 'Fit' slot is going to be removed.

But I would suggest changing the whole tibble section to this:

tibble(
    converged = lavInspect(x, "converged"),
    estimator = lavInspect(x, "options")$estimator,
    ngroups = lavInspect(x, "ngroups"),
    missing_method = lavInspect(x, "options")$missing,
    nobs = sum(purrr::accumulate(lavInspect(x, "nobs"), sum)),
    norig = sum(purrr::accumulate(lavInspect(x, "norig"), sum)),
    nexcluded = norig - nobs
)

avoiding any access to internal slots.

Finally, I don't understand why the purrr::accumulate() function is needed. I believe that what is intended is simply:

nobs = sum(lavInspect(x, "nobs"),
norig = sum(lavInspect(x, "norig"),
nexcluded = norig - nobs

taking the sum over the (possibly multiple) groups.

Yves.

[alexpghayes]

Wonderful, thank you, will make these updates!

[alexpghayes]

@yrosseel Would you be willing to export tidy.lavaan() and glance.lavaan() from lavaan itself, rather than having them live in broom?

[yrosseel]

Hm. No. Because I want to avoid adding additional package dependencies (eg tibble, dplyr, ...).

[simonpcouch]

Hello! Working through this update right now. It looks like the results from purrr::accumulate() and your proposal are indeed different, though:

library(lavaan)
#> This is lavaan 0.6-6
#> lavaan is BETA software! Please report any bugs.
library(broom)
library(tidyverse)

cfa.fit <- cfa(
  'F =~ x1 + x2 + x3 + x4 + x5',
  data = HolzingerSwineford1939, group = "school"
)

# version with purrr
glance.lavaan <- function(x, ...) {
  x %>%
    lavaan::fitmeasures(
      fit.measures =
        c(
          "npar",
          "chisq",
          "rmsea",
          "rmsea.ci.upper",
          "srmr",
          "aic",
          "bic",
          "tli",
          "agfi",
          "cfi"
        )
    ) %>%
    tibble::enframe(name = "term") %>%
    pivot_wider(id_cols = term, names_from = term, values_from = value) %>%
    select(order(colnames(.))) %>%
    map_df(as.numeric) %>%
    bind_cols(
      tibble(
        converged = lavInspect(x, "converged"),
        estimator = lavInspect(x, "options")$estimator,
        ngroups = lavInspect(x, "ngroups"),
        missing_method = lavInspect(x, "options")$missing,
        nobs = sum(purrr::accumulate(lavInspect(x, "nobs"), sum)),
        norig = sum(purrr::accumulate(lavInspect(x, "norig"), sum)),
        nexcluded = norig - nobs
      )
    ) %>%
    rename(rmsea.conf.high = rmsea.ci.upper, AIC = aic, BIC = bic)
}

old <- glance(cfa.fit)

# version with only sum
glance.lavaan <- function(x, ...) {
  x %>%
    lavaan::fitmeasures(
      fit.measures =
        c(
          "npar",
          "chisq",
          "rmsea",
          "rmsea.ci.upper",
          "srmr",
          "aic",
          "bic",
          "tli",
          "agfi",
          "cfi"
        )
    ) %>%
    tibble::enframe(name = "term") %>%
    pivot_wider(id_cols = term, names_from = term, values_from = value) %>%
    select(order(colnames(.))) %>%
    map_df(as.numeric) %>%
    bind_cols(
      tibble(
        converged = lavInspect(x, "converged"),
        estimator = lavInspect(x, "options")$estimator,
        ngroups = lavInspect(x, "ngroups"),
        missing_method = lavInspect(x, "options")$missing,
        nobs = sum(lavInspect(x, "nobs")),
        norig = sum(lavInspect(x, "norig")),
        nexcluded = norig - nobs
      )
    ) %>%
    rename(rmsea.conf.high = rmsea.ci.upper, AIC = aic, BIC = bic)
}

new <- glance(cfa.fit)

old$nobs == new$nobs
#> [1] FALSE
old$norig == new$norig
#> [1] FALSE

^{Created on 2020-05-25 by the reprex package (v0.3.0.9001)}

Will be pushing some changes and referencing this issue here soon.🙂

[yrosseel]

Thanks! The purrr::accumulate() version was always wrong. So the 'new' version is correct!

[simonpcouch]

@benwhalley, any thoughts on nobs and norig representing the cumulative sum rather than just the sum of the groups? Just tagging you as you seem to be the implementer in #5932820.🙂

[benwhalley]

I don't think I did implement that, but it's been a while if I did. No strong views on it now!

[simonpcouch]

Gotcha! Making the switch now, thanks.

[github-actions]

This issue has been automatically locked. If you believe you have found a related problem, please file a new issue (with a reprex: https://reprex.tidyverse.org) and link to this issue.