gtsummary::tbl_regression use pool_and_tidy_mice() with tidy

I'm currently trying to run tbl_regression with an imputed dataset from mice run through a logistic glm. I'm having trouble trying to combine the custom tidiers pool_and_tidy_mice and tidy_standardize in order to get the regression output from the pooled mice results and the standardized odds ratio estimates.

Is there a way to get the standardized odds ratios with pooling imputed data with tbl_regression or possibly another step I can take to get them?

Using the surival package as an example, I can get standardized odds ratio with non-imputed data with this code:

library(tidyverse)
library(broom.mixed)
library(broom)
library(survival)

mod <- glm(death ~ marker   grade, data = trial, family = "binomial")
tbl_regression(mod, tidy_fun = tidy_standardize, 
                    exponentiate = TRUE, conf.int = TRUE, conf.level = 0.95)

But, if I try to use tidy_standardize in the following code,

suppressWarnings(mice::mice(trial, m = 2)) %>%
  with(glm(death ~ marker   grade, family = "binomial")) %>%
  tbl_regression(tidy_fun = tidy_standardize, 
                 exponentiate = TRUE, conf.int = TRUE, conf.level = 0.95)

I get this error:

x There was an error calling `tidy_fun()`. Most likely, this is because the
function supplied in `tidy_fun=` was misspelled, does not exist, is not
compatible with your object, or was missing necessary arguments (e.g. `conf.level=` or `conf.int=`). See error message below.
Error: Error in .model_parameters_generic(model = mice::pool(model), ci = ci, : formal argument "standardize" matched by multiple actual arguments
In addition: Warning messages:
1: Could not get model data. 
2: No variables could be standardized. 
3: Could not get model data. 
4: No variables could be standardized.

I also tried the following but still got an error message:

suppressWarnings(mice::mice(trial, m = 2)) %>%
  with(glm(death ~ marker   grade, family = "binomial")) %>%
  tbl_regression(tidy_fun = purrr::partial(tidy_standardize, method = "posthoc"), 
                 exponentiate = TRUE, conf.int = TRUE, conf.level = 0.95)

x There was an error calling `tidy_fun()`. Most likely, this is because the
function supplied in `tidy_fun=` was misspelled, does not exist, is not
compatible with your object, or was missing necessary arguments (e.g. `conf.level=` or `conf.int=`). See error message below.
Error: Error in .model_parameters_generic(model = mice::pool(model), ci = ci, : formal argument "standardize" matched by multiple actual arguments

CodePudding user response：

Unlike the unimputed data, mice doesn't output a table, so it must be transformed using complete afterwards. What about this:

library(tidyverse)
library(broom.mixed)
library(broom)
library(survival)
library(gtsummary)
library(mice)

data(trial)

mice(trial, m = 2) %>%
  complete() %>%
  as_tibble() %>%
  glm(death ~ marker   grade, data = ., family = "binomial") %>%
  tbl_regression(
    tidy_fun = tidy_standardize,
    exponentiate = TRUE, conf.int = TRUE, conf.level = 0.95
  )

CodePudding user response：

Unfortunately, tidiers cannot be combined or stacked.

To get what you're after, I would recommend that you standardize the covariates in the modelling step, rather than waiting until the model has been estimated.

Example below!

library(gtsummary)
packageVersion("gtsummary")
#> [1] '1.5.0'

tbl <-
  suppressWarnings(mice::mice(trial, m = 2)) %>%
  with(glm(death ~ scale(marker)   grade, family = "binomial")) %>%
  tbl_regression(exponentiate = TRUE)
#> pool_and_tidy_mice(): Tidying mice model with
#> `mice::pool(x) %>% mice::tidy(exponentiate = TRUE, conf.int = TRUE, conf.level = 0.95)`

^{Created on 2021-12-10 by the reprex package (v2.0.1)}