method = "t.test", compare_means calculated P value is different from stat_compare_means

[wydty]

use data

> data("ToothGrowth")
> df <- ToothGrowth
> compare_means(len~dose,data = df,method = "t.test")
# A tibble: 3 x 8
    .y. group1 group2            p        p.adj p.format p.signif method
  <chr>  <chr>  <chr>        <dbl>        <dbl>    <chr>    <chr>  <chr>
1   len    0.5      1 6.697250e-09 1.339450e-08  6.7e-09     **** T-test
2   len    0.5      2 1.469534e-16 4.408602e-16  < 2e-16     **** T-test
3   len      1      2 1.442603e-05 1.442603e-05  1.4e-05     **** T-test
## 0.5 VS 1  p.value=6.697250e-09

> my_comparisons <- list( c("0.5", "1"), c("1", "2"), c("0.5", "2") )
> ggboxplot(df, x = "dose", y = "len")+ 
+  stat_compare_means(comparisons = my_comparisons,method = "t.test")
## 0.5 VS 1  p.value≈1.3e-07

#t.test result  0.5 VS 1
> t.test(df$len[df$dose==0.5],y=df$len[df$dose==1])$p.value
[1] 1.268301e-07

t.test() calculated P value the same of graphics，but compare_means() function is different

[kassambara]

Thank you for pointing this out.

This is due to inconsistency in R base functions.

Internally, compare_means() uses the R base function pairewise.t.test(), which default is to compute pairewise t.test by assumming homocedacity.

In stat_compare_means(), if the argument comparisons is specified, the standard t.test() function is used to compare only the groups specified by users. The problem is that, the default behavior of the standard t.test() is to not assume homocedacity.

This issue is fixed. Now, the default behaviour of pairewise.t.test() is to act as the standard t.test() for consistency in ggpubr.

You can install the latest developmental version of ggpubr (devtools::install_github("kassambara/ggpubr").

Kindly confirm that it works so that we can close the issue, thanks.

[wydty]

I see,Thank you very much

[kcmtest]

I have similar issue with the KW test here is my code and the output of the same ,when im using the base R function the p value reported is different im using the latest version Version: 0.2.999
not sure how to add the generated plot as don't know in github yet but you can run the code and data to see what im doing wrong

My data so that you can run and test to see whats wrong https://drive.google.com/open?id=10vanPZEW77qfImINZvoYiAc8CH2Y0XOi

library(ggpubr)
library(reshape2)
df1 = read.csv("NEW_RBP/CORPLOT/RBP_DISEASE_CLUSTER/RBP_DISEASE_C1_C4_AVG.txt",header = TRUE,sep = "\t",row.names = 1)
head(df1)
my_comparisons <- list( c("HSC", "LSC"), c("LSC", "Blast"), c("HSC", "Blast") )
 
ex <- melt(df1, id.vars=NULL)

head(ex)

ggboxplot(ex, x = "variable", y = "value",ggtheme = theme_bw(base_size = 30),
          color = "variable", palette = "jco", add = "jitter")+ 
  stat_compare_means(comparisons = my_comparisons)+ # Add pairwise comparisons p-value
  stat_compare_means(label.y = 50,method = "wilcox.test")     # Add global p-value


kruskal.test(LSC ~ HSC, data = df1)

	Kruskal-Wallis rank sum test

data:  LSC by HSC
Kruskal-Wallis chi-squared = 416, df = 416, p-value = 0.4908