plot_predictions() with categorical variables and vcov = FALSE results in error message

[huftis]

Using plot_prediction() with categorical variables and vcov = FALSE results in an error message (setting vcov = TRUE works fine):

library(marginaleffects)
mod = lm(Sepal.Length ~ Species, data = iris)
plot_predictions(mod, by = "Species", vcov = FALSE)
#> Error in `ggplot2::geom_point()`:
#> ! Problem while computing aesthetics.
#> ℹ Error occurred in the 1st layer.
#> Caused by error in `.data[[NULL]]`:
#> ! Must subset the data pronoun with a string, not `NULL`.

Tested using the latest Git version.

This is especially a problem for glmmTMB models, where plot_predictions() tells us that we should use vcov = FALSE (though it actually seems to work fine with vcov = vcov(mod) for linear Gaussian models like this one).

[vincentarelbundock]

Thanks a lot for the report. I think that the problem arises here, where we call .data[[v_color]] but v_color is NULL:

https://github.com/vincentarelbundock/marginaleffects/blob/main/R/plot_build.R#L79

It's probably just a question of inserting more if/else statements or use a do.call() to ensure that we don't feed useless arguments to ggplot::aes(). The problem will probably arise in the next code block as well with continuous x-axis, but it should be a pretty easy fix. (But start of semester is upon us at my university, so I'll do my best...)

[vincentarelbundock]

Half-baked PR: #898

[vincentarelbundock]

Thanks again for the report. Should be fixed on Github:

library(marginaleffects)
mod = lm(Sepal.Length ~ Species, data = iris)
plot_predictions(mod, by = "Species", vcov = FALSE)