CODEWITHSUNDEEP
rgtsummary
rhysticlight
rhysticlight

Reputation: 31

Problem modifying significant digits in gtsummary confidence intervals

I am using the tbl_summary() function from the gtsummary package to summarize several continuous variables in my data set. I would like to print a 95% confidence interval for the mean and am using the add_ci() function to accomplish this. The problem is that I cannot get the column to print the appropriate number of significant digits for the CI.

df %>% select("dcs.pre.total", "dcs.post.total") %>% 
  tbl_summary(percent = "row",
            type = everything() ~ "continuous",
            statistic = list(all_continuous () ~ "{mean}"),
            digits = all_continuous () ~ 2) %>%
  add_ci(pattern = "{stat} ({ci})",
         style_fun = everything() ~ style_sigfig)

This produces a mean value with two digits after the decimal place, but only rounded whole numbers for the CI.

I have tried:

df %>% select("dcs.pre.total", "dcs.post.total") %>% 
  tbl_summary(percent = "row",
            type = everything() ~ "continuous",
            statistic = list(all_continuous () ~ "{mean}"),
            digits = all_continuous () ~ 2) %>%
  add_ci(pattern = "{stat} ({ci})",
         style_fun = everything() ~ style_sigfig(digits = 2))

But get this error:

Error in eval_tidy(pair$lhs, env = default_env) : 
  argument "x" is missing, with no default

Any help is greatly appreciated.

Edit, sample of data:

dput(head(df[,c(83, 84)]))
structure(list(dcs.pre.total = c(43.75, 70.8333333333333, 68.75, 
75, 64.5833333333333, 56.25), dcs.post.total = c(25, 4.16666666666667, 
16.6666666666667, 16.6666666666667, 20.8333333333333, 10.4166666666667
)), row.names = c(2L, 3L, 4L, 5L, 6L, 9L), class = "data.frame")

Upvotes: 3

Views: 1366

Answers (1)

TrainingPizza
TrainingPizza

Reputation: 1150

Thanks for the sample data. You can use purrr::partial for this task. Rather than use style_sigfig, it's also more appropriate to use style_number. With style_sigfig, values up to the thousandth will be displayed for the lower bound of dcs.post.total in your sample data as 4 digits will always be displayed with digits = 4. It may be that style_sigfig is fine for your actual data, but style_number is generally more applicable.

library(gtsummary)
df <-
  structure(list(dcs.pre.total = c(43.75, 70.8333333333333, 68.75, 
  75, 64.5833333333333, 56.25), dcs.post.total = c(25, 4.16666666666667, 
  16.6666666666667, 16.6666666666667, 20.8333333333333, 10.4166666666667
  )), row.names = c(2L, 3L, 4L, 5L, 6L, 9L), class = "data.frame")


df %>% select("dcs.pre.total", "dcs.post.total") %>% 
  tbl_summary(percent = "row",
            type = everything() ~ "continuous",
            statistic = list(all_continuous () ~ "{mean}"),
            digits = all_continuous () ~ 2) %>%
  add_ci(pattern = "{stat} ({ci})",
         style_fun = everything() ~ purrr::partial(style_number, digits = 2))

Upvotes: 3

Related Questions