Bug Report: as_flex_table() error using p-value from svy.wald.test

[jmbarajas]

Hi there,

I'm encountering an error when converting a tbl_svysummary object to a flextable using as_flex_table() when it has a p-value from the svy.wald.test. This seems to have broken between release 1.5.0 and 1.5.1. See reprexes below: the first code chunk works when outputting to html.

library(survey)
#> Loading required package: grid
#> Loading required package: Matrix
#> Loading required package: survival
#> 
#> Attaching package: 'survey'
#> The following object is masked from 'package:graphics':
#> 
#>     dotchart
library(gtsummary)

data(api)
dstrat<-svydesign(id=~1,strata=~stype, weights=~pw, data=apistrat, fpc=~fpc)

tbl_svysummary(dstrat,
               by = awards,
               include = c(api00, api99)) %>% 
  add_p(test = list(all_continuous() ~ "svy.wald.test"))

Characteristic	No, N = 2,2361	Yes, N = 3,9581	p-value2
api00	641 (528, 743)	671 (588, 765)	<0.001
api99	636 (524, 724)	625 (526, 727)	<0.001
1 Median (IQR)
2 Wald test of independence for complex survey samples

^{Created on 2022-02-15 by the reprex package (v2.0.1)}

But the second fails when converting.

library(survey)
#> Loading required package: grid
#> Loading required package: Matrix
#> Loading required package: survival
#> 
#> Attaching package: 'survey'
#> The following object is masked from 'package:graphics':
#> 
#>     dotchart
library(gtsummary)

data(api)
dstrat<-svydesign(id=~1,strata=~stype, weights=~pw, data=apistrat, fpc=~fpc)

tbl_svysummary(dstrat,
               by = awards,
               include = c(api00, api99)) %>% 
  add_p(test = list(all_continuous() ~ "svy.wald.test")) %>% 
  as_flex_table()
#> Error:
#> ! Assigned data `df_updated` must be compatible with existing data.
#> i Error occurred for column `p.value`.
#> x Can't convert <character[,1]> to <character>.
#> Cannot decrease dimensions.

^{Created on 2022-02-15 by the reprex package (v2.0.1)}

The error doesn't occur with the default Wilcoxon rank sum test. The code appears to be choking because the p-value is stored as matrix rather than a single numeric value, but I don't know what change was introduced to cause the bug.

[ddsjoberg]

Thank you for letting me know! The p-value column was being saved as a nX1 matrix in the data frame, as opposed to a numeric vector. I'll need to fix that, but in the meantime, you can use the code below to get a working table.

library(gtsummary)

data(api, package = "survey")
dstrat <- survey::svydesign(id=~1,strata=~stype, weights=~pw, data=apistrat, fpc=~fpc)

tbl <- 
  tbl_svysummary(
    dstrat,
    by = awards,
    include = c(api00, api99)
  ) %>% 
  add_p(test = list(all_continuous() ~ "svy.wald.test")) %>%
  # convert these columns to numeric vectors (somehow was stored as a nX1 matrix)
  modify_table_body(
    ~.x %>% dplyr::mutate(dplyr::across(c(statistic, p.value), as.numeric))
  )

as_flex_table(tbl)

^{Created on 2022-02-15 by the reprex package (v2.0.1)}

[ddsjoberg]

Issue submitted to the broom package tidymodels/broom#1081

[jmbarajas]

Thanks for the workaround and upcoming fix!