p-values from compare_means() and wilcox.test() are different

[smshuai]

Hello,

Below is an example:

val = c(-0.78137127,-0.86180992,-0.91177614,-0.95413924,-0.80979775,
        -0.70236469,-0.96355688,-0.84418155,-0.30040466,1.25324304,
        0.53833376,3.01788826,5.35022326)
grp = c(rep('G1', 8), rep('G2', 5))
df = data.frame(val, grp)
wilcox.test(val~grp, df)
require(ggpubr)
compare_means(val~grp, df)

The p-value from wilcox.test is 0.001554 while from compare_menas is 0.00431.

Any clue about the cause? Thanks!

Session info:

> sessionInfo()
R version 3.4.1 (2017-06-30)
Platform: x86_64-apple-darwin15.6.0 (64-bit)
Running under: macOS High Sierra 10.13.6

Matrix products: default
BLAS: /System/Library/Frameworks/Accelerate.framework/Versions/A/Frameworks/vecLib.framework/Versions/A/libBLAS.dylib
LAPACK: /Library/Frameworks/R.framework/Versions/3.4/Resources/lib/libRlapack.dylib

locale:
[1] en_CA.UTF-8/en_CA.UTF-8/en_CA.UTF-8/C/en_CA.UTF-8/en_CA.UTF-8

attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] bindrcpp_0.2.2 ggpubr_0.2     magrittr_1.5   ggplot2_3.0.0 

loaded via a namespace (and not attached):
 [1] Rcpp_0.12.17     rstudioapi_0.7   bindr_0.1.1      tidyselect_0.2.4 munsell_0.5.0   
 [6] colorspace_1.3-2 R6_2.2.2         rlang_0.2.1      plyr_1.8.4       dplyr_0.7.6     
[11] tools_3.4.1      grid_3.4.1       gtable_0.2.0     utf8_1.1.4       cli_1.0.0       
[16] withr_2.1.2      yaml_2.1.19      lazyeval_0.2.1   assertthat_0.2.0 tibble_1.4.2    
[21] crayon_1.3.4     purrr_0.2.5      tidyr_0.8.1      glue_1.2.0       compiler_3.4.1  
[26] pillar_1.2.3     scales_0.5.0     pkgconfig_2.0.1 

[TapsiS]

I am also having different p values. Here is an example -

# example dataframe
before <-c(200.1, 190.9, 192.7, 213, 241.4, 196.9, 172.2, 185.5, 205.2, 193.7)
after <-c(392.9, 393.2, 345.1, 393, 434, 427.9, 422, 383.9, 392.3, 352.2)

# Create a data frame
my_data <- data.frame( 
        group = rep(c("before", "after"), each = 10),
        weight = c(before,  after)
        )
# standard wilcox test
wilcox.test(after, before, paired = TRUE, alternative = "two.sided")

##V = 55, p-value = 0.001953

# using stat_compare_means
ggboxplot(my_data, x = "group", y = "weight",
                                         color  = "group", palette =c("#00AFBB", "#E7B800"),
                                         add = "jitter", shape = "group") +
           stat_compare_means(comparisons = list( c("after", "before")), method = "wilcox.test", paired = TRUE,
                                      method.args = list(alternative = "two.sided"))

##p value 0.0059

[stemicha]

same here:
library(ggpubr)
library(tidyverse)

set.seed(666)

#genrate data
df <- tibble(group=c(rep("group1",5),rep("group2",5)),int=c(rnorm(n = 5,mean = 100),rnorm(5,mean = 1000)))
#log data
df$log2_int <- log2(df$int)

#on log2 data
#wilcox test
wilcox.test(x = unlist(df %>% filter(group=="group1") %>% select(log2_int)),
y = unlist(df %>% filter(group=="group2") %>% select(log2_int)),
paired = F,
alternative = "two.sided")

p-value = 0.007937

#ggpubr
compare_means(log2_int ~ group, df,method = "wilcox.test",paired = F,alternative = "two.sided")

p-value = 0.012

[kassambara]

Fixed now, thanks!

When method = "wilcox.test", the function compare_means() set automatically the option exact = FALSE. This is no longer the case

[kazumits]

Please also remove the exact=FALSE in stat_compare_means().

[kassambara]

removed now, thanks

[hitrp]

I found there're different results for wilcox.test(df,mu=xx) and compare_means(df,ref.group='.all.') for one sample test, after digging into the details, I found that compare_means seems to take all values in df as .all. group, and operate a two-sample test(wilcox rank sum) rather than one sample test, which may violate the independent assumption, am I think it in the right way?

Here's an easy one to repeat the results:

val = c(-0.78137127,-0.86180992,-0.91177614,-0.95413924,-0.80979775,
-0.70236469,-0.96355688,-0.84418155,-0.30040466,1.25324304,
0.53833376,3.01788826,5.35022326)
grp = c(rep('G1', 8), rep('G2', 5))
df = data.frame(val, grp)
wilcox.test(df[df['grp']=='G1','val'], mu=mean(df[,'val']))
compare_means(val~grp, df, ref.group='.all.')

Also, this is the detail of compare_means() function for handing '.all.'
grp val .group.
1 G1 -0.7813713 G1
2 G1 -0.8618099 G1
3 G1 -0.9117761 G1
4 G1 -0.9541392 G1
5 G1 -0.8097977 G1
6 G1 -0.7023647 G1
7 G1 -0.9635569 G1
8 G1 -0.8441815 G1
9 G2 -0.3004047 G2
10 G2 1.2532430 G2
11 G2 0.5383338 G2
12 G2 3.0178883 G2
13 G2 5.3502233 G2
14 G1 -0.7813713 .all.
15 G1 -0.8618099 .all.
16 G1 -0.9117761 .all.
17 G1 -0.9541392 .all.
18 G1 -0.8097977 .all.
19 G1 -0.7023647 .all.
20 G1 -0.9635569 .all.
21 G1 -0.8441815 .all.
22 G2 -0.3004047 .all.
23 G2 1.2532430 .all.
24 G2 0.5383338 .all.
25 G2 3.0178883 .all.
26 G2 5.3502233 .all.
val ~ .group.