add_p() doesn't work for difftime values #880

oranwutang · 2021-04-29T10:00:06Z

Hi!

I have detected that add_p() doesn't work with columns of class difftime. As a matter of fact it is related with wilcox.test() which doesn't accept that class, but if the variable is parsed to numeric class with as.numeric() before using it on gtsummary workflow, it does work.

So my suggestion is that add_p() should check if variables are of difftime class and in that case parse them with as.numeric() before inputting them into wilcox.test().

PS: t.test() does accept difftime objects

This issue is somewhat related with (#343).

Thanks in advance!

(Sorry for the horrible reprex which is rendered with the html code included, but at least the tables are rendered)

library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union
library(lubridate)
#> 
#> Attaching package: 'lubridate'
#> The following objects are masked from 'package:base':
#> 
#>     date, intersect, setdiff, union
library(gtsummary)

set.seed(123)

time_diff <- lubridate::as.difftime(rnorm(250, 2000, 100), units = "mins")
group_sample <- sample(c("A", "B"), 250, replace = TRUE)

df <- data.frame(group_sample, time_diff)

df %>% select(group_sample, time_diff) %>% 
  tbl_summary(by=group_sample) %>% add_p()
#> There was an error in 'add_p()/add_difference()' for variable 'time_diff', p-value omitted:
#> Error in wilcox.test.default(x = structure(c(2012.92877351609, 2171.50649868833, : 'x' must be numeric

Characteristic	A, N = 118¹	B, N = 132¹	p-value
time_diff	1,992 (1,934, 2,063)	1,981 (1,938, 2,055)
¹ `Median (IQR) <br /> </p> </td> </tr>`

class(df$time_diff)
#> [1] "difftime"

t.test(df$time_diff~df$group_sample)
#>
#> Welch Two Sample t-test
#>
#> data: df$time_diff by df$group_sample
#> t = 0.38027 mins, df = 237.43, p-value = 0.7041 mins
#> alternative hypothesis: true difference in means is not equal to 0
#> 95 percent confidence interval:
#> -19.11303 mins 28.25683 mins
#> sample estimates:
#> Time differences in mins
#> mean in group A mean in group B
#> 2001.558 1996.986

wilcox.test(df$time_diff~df$group_sample)
#> Error in wilcox.test.default(x = structure(c(2012.92877351609, 2171.50649868833, : 'x' must be numeric

df$time_diff <- as.numeric(df$time_diff)

df %>% select(group_sample, time_diff) %>%
tbl_summary(by=group_sample) %>% add_p()

Characteristic	A, N = 118¹	B, N = 132¹	p-value²
time_diff	1,992 (1,934, 2,063)	1,981 (1,938, 2,055)	0.9
¹ `Median (IQR) <br /> </p> <p class="gt_footnote"> <sup class="gt_footnote_marks"> <em>2</em> </sup> Wilcoxon rank sum test <br /> </p> </td> </tr>`

^{Created on 2021-04-29 by the reprex package (v0.3.0)}

The text was updated successfully, but these errors were encountered:

ddsjoberg · 2021-04-29T13:08:54Z

Hey @oranwutang ! Can you review this pull request? #881
There is a checklist of items to review :)

library(gtsummary)

set.seed(123)

time_diff <- lubridate::as.difftime(rnorm(250, 2000, 100), units = "mins")
group_sample <- sample(c("A", "B"), 250, replace = TRUE)

df <- data.frame(group_sample, time_diff)

df %>% 
  select(group_sample, time_diff) %>% 
  tbl_summary(by=group_sample) %>%
  add_p() %>%
  as_kable()

Characteristic	A, N = 118	B, N = 132	p-value
time_diff	1,992 (1,934, 2,063)	1,981 (1,938, 2,055)	0.9

^{Created on 2021-04-29 by the reprex package (v2.0.0)}

ddsjoberg mentioned this issue Apr 29, 2021

difftime wilcox.test() update #881

Merged

14 tasks

ddsjoberg closed this as completed in #881 Apr 29, 2021

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

add_p() doesn't work for difftime values #880

add_p() doesn't work for difftime values #880

oranwutang commented Apr 29, 2021

ddsjoberg commented Apr 29, 2021

add_p() doesn't work for difftime values #880

add_p() doesn't work for difftime values #880

Comments

oranwutang commented Apr 29, 2021

ddsjoberg commented Apr 29, 2021