redirected from r-lib/rlang :sim_slopes encounters an issue with rlang-package

[Jendryczko]

Hello everybody.

I already reported this problem to r-lib/rlang:
r-lib/rlang#1529

but I was asked to report it here, so I repeat it:

It seems that there is some sort of compatability issue with the latest version of the interactions-package with the latest version of the rlang-package, and -potentially - also the latest version of R and RStudio. If I run the first example provided for "sim_slopes":

fiti <- lm(Income ~ Frost + Murder * Illiteracy, data = as.data.frame(state.x77))
sim_slopes(model = fiti, pred = Murder, modx = Illiteracy)

I get the following error:
Error:
! class must be a character vector, not a list.
Backtrace:

interactions::sim_slopes(model = fiti, pred = Murder, modx = Illiteracy)
jtools:::summ.lm(...)
jtools:::do_robust(model, robust, cluster, data, vcov)
jtools::get_robust_se(...)
jtools::stop_wrap(...)
rlang (local) <fn>(class = )

versions:
packageVersion('interactions')# ‘1.1.5'
packageVersion('rlang')# ‘1.0.6’
packageVersion('jtools')# 2.2.1
#R-version: 4.2.2.
#RStudio-version: 2022.07.2 Build 576

There already exists a stackoverflow entry for this:
https://stackoverflow.com/questions/74720514/interactions-and-rlang-error-class-must-be-a-character-vector-not-a-list

But updating the dependencies for jtools and rlang did not help in my case. Does anybody have any solutions?

[jacob-long]

I've been unable to replicate this, so I'm fairly puzzled. Can you (and any others experiencing the error) give the full output of sessionInfo()?

[Jendryczko]

Thank you for attending to this problem. Here you go:

library(interactions)

Using a fitted model as formula input

fiti <- lm(Income ~ Frost + Murder * Illiteracy,
data=as.data.frame(state.x77))
probe_interaction(model = fiti, pred = Murder, modx = Illiteracy,
modx.values = "plus-minus")

sessionInfo()

Output:

Error:
! class must be a character vector, not a list.
Backtrace:

1. interactions::probe_interaction(...)
2. interactions::sim_slopes(...)
3. jtools:::summ.lm(...)
4. jtools:::do_robust(model, robust, cluster, data, vcov)
5. jtools::get_robust_se(...)
6. jtools::stop_wrap(...)
7. rlang (local) <fn>(class = )

sessionInfo()
R version 4.2.2 (2022-10-31 ucrt)
Platform: x86_64-w64-mingw32/x64 (64-bit)
Running under: Windows 10 x64 (build 19044)

Matrix products: default

locale:
[1] LC_COLLATE=German_Germany.utf8 LC_CTYPE=German_Germany.utf8 LC_MONETARY=German_Germany.utf8
[4] LC_NUMERIC=C LC_TIME=German_Germany.utf8

attached base packages:
[1] stats graphics grDevices utils datasets methods base

other attached packages:
[1] interactions_1.1.5

loaded via a namespace (and not attached):
[1] Rcpp_1.0.9 rstudioapi_0.14 magrittr_2.0.3 tidyselect_1.2.0 munsell_0.5.0 colorspace_2.0-3
[7] R6_2.5.1 rlang_1.0.6 fansi_1.0.3 dplyr_1.0.10 tools_4.2.2 grid_4.2.2
[13] gtable_0.3.1 utf8_1.2.2 cli_3.5.0 DBI_1.1.3 digest_0.6.31 assertthat_0.2.1
[19] tibble_3.1.8 lifecycle_1.0.3 crayon_1.5.2 ggplot2_3.4.0 vctrs_0.5.1 jtools_2.2.1
[25] glue_1.6.2 compiler_4.2.2 pander_0.6.5 pillar_1.8.1 generics_0.1.3 scales_1.2.1
[31] pkgconfig_2.0.3

[zeileis]

The problem only occurs if the sandwich package is not installed. Thus, @Jendryczko, a quick fix is to install the sandwich package.

What happens internally is the following:

• sim_slopes() calls get_robust_se()
• get_robust_se() checks whether the sandwich package is available
• if not, it tries to throw an informative error that sandwich is needed via: stop_wrap("When using robust SEs you need to have the \'sandwich\' package.", call. = FALSE).
• The call. = FALSE argument is not caught by jtools::stop_wrap() but passed on to rlang which then fails

My guess is that jtools::stop_wrap() needs to handle the class. = FALSE argument correctly , similar to jtools::warn_wrap(). Or possibly the stop_wrap() calls in the packages need to be fixed. @jacob-long, it would be great if you could have a look at this.

[Jendryczko]

Thank you very much @zeileis. You were right, simply installing the sandwich-package (no need to load it with library()) fixed everything. I could swear that when I installed ineractions in older R-versions, sandwich was automatically installed alongside but I may be remembering that wrong.

[zeileis]

Possibly you indicated that you want all dependencies to be installed (and not just the required Depends/Imports). The sandwich package is only a "Suggests" and not an "Imports" dependency.

[HUtangge]

Thank you so much :)