plot_predictions() won't accept mida object

[mattjamesmartin]

Bug reports must include:

1. I am running a logit model using data imputed from the mice package. I am trying to then use plot_predictions() to visualize my pooled results, but it is not working seemingly due to a bug. It won't accept the mida object produced by my model. Below is an example using mtcars:

2. 
   data(mtcars)
   set.seed(123)

Create random missing values

mtcars_with_missing <- mtcars missing_prop <- 0.1 # Proportion of missing values n_missing <- round(nrow(mtcars) * missing_prop)
rows_to_replace <- sample(1:nrow(mtcars), size = n_missing, replace = FALSE)

Replace values with NA

mtcars_with_missing[rows_to_replace, ] <- NA

Impute data

mtcars_imp <- mice(mtcars_with_missing, m = 5, method = "pmm", seed = 123)

Create mida object

mtcars_imp_logit <- with(data = mtcars_imp, exp = glm(am ~ cyl + hp + wt, family = "binomial"))

plot_predictions(mtcars_imp_logit, condition = "hp")

4. Output: `plot_predictions(mtcars_imp_logit, condition = "hp") Error: Entries in the "condition" argument must be element of:`

[vincentarelbundock]

Try specifying the type and newdata arguments explicitly:

plot_predictions(
    mtcars_imp_logit, 
    condition = "hp", 
    type = "response",
    newdata = mtcars_with_missing)

[mattjamesmartin]

That works! Thank you so much for your fast reply.