CODEWITHSUNDEEP
gtsummary
Uknowepi
Uknowepi

Reputation: 53

How to use car::Anova in tbl_uvregression?

I am trying to use a gee like the following code:

example <- trial%>%
  na.omit()%>%
  tbl_uvregression(
    method = geepack::geeglm,
    y = death,
    method.args = list(
      family = binomial,
      id = grade,
      corstr = "independence"),
    exponentiate = TRUE,
    add_estimate_to_reference_rows = T,
    conf.level = 0.95)

Which works just fine and gives me desired univariate analysis for all Xs. However, when I try to use add_global_p, I get the following error: There was an error running car::Anova() for "trt", likely due to this model type not being supported. The results displayed are based on add_global_p(anova_fun = gtsummary::tidy_wald_test). This also happens to all Xs. However, I know for sure this model is supported by car::Anova since I used it with geeglm (see example 1 down):

example1 <- trial%>%na.omit()

example1.outcome <- geepack::geeglm(death ~ trt,family = binomial,id = grade,corstr = "independence", data = example1)
car::Anova(example1.outcome, type = 3, test.statistic = "Wald")

Outcome:

Analysis of Deviance table (Type III tests) Response: death

variable Df Chisq Pr(>chisq)
intercept 1 0.28 0.6
trt 1 2.03 0.15

Going back to example, I tried a few different things such as:

example%>%
    add_global_p(car::Anova(type = 3, test.statistic = "Chisq"))

But it gives the following error: Error in car::Anova(type = 3, test.statistic = "Chisq") : argument "mod" is missing, with no default.

I also tried:

add_global_p(car::Anova(example, type = 3, test.statistic = "Chisq"))

but the following error arises: Error in UseMethod("vcov") : no applicable method for 'vcov' applied to an object of class "c('tbl_uvregression', 'gtsummary')"

I also tried:

example%>%
    add_global_p(anova_fun= car::Anova)

But it says argument mod is missing. I tried a few different things, but I do not seem to be able to get it. I also looked at the following question, and still could not figure it out: Problem using 'anova_fun' argument in 'add_global_p()'

Bonus: I am also wondering if it would be possible to add the values of Chisq test (the 3rd column in the output of example 1) in a column next to p-values.

Upvotes: 0

Views: 250

Answers (1)

Daniel D. Sjoberg
Daniel D. Sjoberg

Reputation: 11774

See below for a working example:

library(gtsummary)

tbl <-
  trial %>%
  na.omit()%>%
  tbl_uvregression(
    method = geepack::geeglm,
    y = death,
    include = response,
    method.args = list(
      family = binomial,
      id = grade,
      corstr = "independence"),
    exponentiate = TRUE,
    add_estimate_to_reference_rows = T,
    conf.level = 0.95
  )

# this is the geepack::geeglm() model
class(tbl$tbls$response$model_obj)
#> [1] "geeglm" "gee"    "glm"    "lm"

# not compatible with `car::Anova()`
car::Anova(tbl$tbls$response$model_obj)
#> Error in glm.control(trace = 0L, jack = 0L, j1s = 0L, fij = 0L, maxit = 25L, : unused arguments (jack = 0, j1s = 0, fij = 0)

# the non-compatibility is shown below from within gtsummary
tbl_error <- tbl %>% add_global_p()
#> ✖ There was an error running `car::Anova()` for "response", likely due to this model type not being supported. The results displayed are based on `add_global_p(anova_fun = gtsummary::tidy_wald_test)`

# as the error message indicates, use the tidy_wald_test() function to get around the error
# FYI, this function was written FOR geepack models
tbl %>%
  add_global_p(anova_fun = gtsummary::tidy_wald_test) %>%
  as_kable() # convert to kable so it will display on StackOverflow
Characteristic N OR 95% CI p-value
Tumor Response 173 0.35 0.19, 0.66 <0.001

Created on 2023-06-29 with reprex v2.0.2

Upvotes: 0

Related Questions