Bug Report: Variable with all NA levels causes error

[szimmer]

Per discussion on Posit Community forum, a error is yielded when a column consists entirely of missing data. This should be handled more gracefully

---

library(tidyverse)
library(gtsummary)

set.seed(8675309)

mydf <- tibble(
  Group=rep(1:4, each=20),
  CatValid=as.factor(sample(letters[1:4], 80, replace=TRUE)),
  CatNA=factor(NA)
)

mydf %>%
  select(-CatNA) %>%
  tbl_summary(by="Group")

Characteristic	1 N = 201	2 N = 201	3 N = 201	4 N = 201
CatValid
a	5 (25%)	3 (15%)	4 (20%)	4 (20%)
b	4 (20%)	8 (40%)	6 (30%)	4 (20%)
c	8 (40%)	4 (20%)	6 (30%)	6 (30%)
d	3 (15%)	5 (25%)	4 (20%)	6 (30%)
1 n (%)

mydf %>%
  tbl_summary(by="Group")
#> Error in names(df) <- repaired_names(c(names2(dimnames(x)), n), repair_hint = TRUE, : 'names' attribute [3] must be the same length as the vector [2]

^{Created on 2024-09-13 with reprex v2.1.0}

[ddsjoberg]

Thank you @szimmer for the clear and minimal reprex!

We had already planned to add better messaging for factors with NA levels, but hadn't noted factors with no levels. This will be updated here FYI insightsengineering/cards#255

Having all NAs is ok when the levels of the factor are available.

library(gtsummary)
set.seed(8675309)

mydf <- dplyr::tibble(
  Group=rep(1:4, each=20),
  CatValid=as.factor(sample(letters[1:4], 80, replace=TRUE)),
  CatNA=factor(NA, levels = c("Hello", "There"))
)

tbl_summary(mydf, by = Group) |> 
  bold_labels() |> 
  as_kable()

Characteristic	1 N = 20	2 N = 20	3 N = 20	4 N = 20
CatValid
a	5 (25%)	3 (15%)	4 (20%)	4 (20%)
b	4 (20%)	8 (40%)	6 (30%)	4 (20%)
c	8 (40%)	4 (20%)	6 (30%)	6 (30%)
d	3 (15%)	5 (25%)	4 (20%)	6 (30%)
CatNA
Hello	0 (NA%)	0 (NA%)	0 (NA%)	0 (NA%)
There	0 (NA%)	0 (NA%)	0 (NA%)	0 (NA%)
Unknown	20	20	20	20

^{Created on 2024-09-13 with reprex v2.1.0}

[ddsjoberg]

NOTE FOR gtsummary DEVS: Once the {cards} update is complete, add a unit test for the messaging in this situation.