Capital letter NS. remaining after hide.ns in ggplot with ggpubr stat_compare_means

[natjoh]

I want to hide.ns in a ggplot after stat_compare_means is added, it hides all but one random capital letter NS.
install.packages("tidyverse")
library(tidyverse)
library(ggpubr)

dat2<-data.frame(R_Full_simple)
R_Full_simple

#ggplot build
dat2$grp.fac <- factor(dat2$spacer_type,levels=c("1","2","0","3","4"),
labels = c("a) non-valved \n spacer", "b) valved \n spacer",
"c) direct", "d) adapter", "e) syringe") )
dat2$grp.size <- factor(dat2$ett_size,levels=c("1","2","3","4"),
labels = c("3 mm", "5 mm", "7 mm", "9 mm"))

ett <- ggplot(dat2, outlier.shape = NA, mapping = aes(x = grp.fac, y = ett_pct, fill = grp.fac))+
facet_wrap(~grp.size, scales = "free_y")+
geom_boxplot()+geom_point()+theme_classic()
ett
#compare within group
ett+stat_compare_means(comparisons = list(c("c) direct", "a) non-valved \n spacer"),
c("a) non-valved \n spacer", "b) valved \n spacer"),
c("b) valved \n spacer", "d) adapter"), c("b) valved \n spacer",
"e) syringe"),
c("d) adapter", "e) syringe"), c("c) direct", "b) valved \n spacer"),
c("c) direct", "d) adapter"), c("c) direct", "e) syringe"),
c("a) non-valved \n spacer", "e) syringe"), c("a) non-valved \n spacer",
"d) adapter")),
label = "p.signif", hide.ns = TRUE)+
stat_compare_means(method = "anova",label.y = -2)

R_Full_simple.xlsx

[Sergei-EMC]

Hi natjoh,

I'm struggling with the same problem and have searched everywhere for a solution but unfortunately found nothing. Did you come up with a solution to this problem? Thanks in advance

Sergei

[kassambara]

Would you please provide your session info? I can't reproduce this issue with current dev version of ggpubr.

library("ggpubr")
comparisons <- list(c("ctrl", "trt1"), c("ctrl", "trt2"), c("trt1", "trt2"))
ggboxplot(PlantGrowth, x = "group", y = "weight") +
  stat_compare_means(comparisons = comparisons, label = "p.signif")

Hide ns:

comparisons <- list(c("ctrl", "trt1"), c("ctrl", "trt2"), c("trt1", "trt2"))
ggboxplot(PlantGrowth, x = "group", y = "weight") +
  stat_compare_means(comparisons = comparisons, label = "p.signif", hide.ns = TRUE)

Anyway, the option hide.ns avoids printing "NS" for non-significant comparisons, but this won't prevent the lines from being drawn.

An alternative solution would be to go as follow:

library(rstatix) # devtools::install("kassambara/rstatix")
library(ggpubr) # devtools::install("kassambara/ggpubr")
stat.test <- PlantGrowth %>% 
  t_test(weight ~ group) %>%
  add_xy_position()
ggboxplot(PlantGrowth, x = "group", y = "weight") +
  stat_pvalue_manual(stat.test, hide.ns = TRUE)

stat.test

# A tibble: 3 x 12
  .y.    group1 group2 statistic    df       p  p.adj p.adj.signif y.position groups     xmin  xmax
  <chr>  <chr>  <chr>      <dbl> <dbl>   <dbl>  <dbl> <chr>             <dbl> <list>    <int> <int>
1 weight ctrl   trt1        1.19  16.5 0.25    0.25   ns                 6.34 <chr [2]>     1     2
2 weight ctrl   trt2       -2.13  16.8 0.0479  0.0958 ns                 6.39 <chr [2]>     1     3
3 weight trt1   trt2       -3.01  14.1 0.00930 0.0279 *                  6.44 <chr [2]>     2     3

[Sergei-EMC]

Hi,

Thank you for your reply! I'm on R version 3.4.1 (2017-06-30), ggpubr_0.2.1 .

If I run this code:

G8 <-ggplot(CC10,aes(x=variable, y=value)) +stat_boxplot(geom = "errorbar", width = 0.25,linetype=1)+ geom_boxplot(fill=c("turquoise4","lightcoral", "steelblue1"),outlier.shape = 1)+scale_fill_brewer(palette="Set1")+ theme_classic()+ scale_x_discrete(breaks=c("H_score_CC10_normal", "H_score_CC10_CPAM", "H_score_CC10_MP"), labels=c("Normal" ,"CPAM", "MP"))+ theme(axis.title.x = element_blank(),axis.title.y = element_blank())+ stat_compare_means(comparisons = list(c("H_score_CC10_normal","H_score_CC10_CPAM"),c("H_score_CC10_CPAM","H_score_CC10_MP"), c("H_score_CC10_normal","H_score_CC10_MP")),paired=TRUE, method="wilcox.test", label="p.signif", hide.ns=TRUE, symnum.args= list(cutpoints = c(0, 0.017, 1), symbols = c("*", "ns")), label.y = c(320,335,350),tip.length=0.02)

I get

`
It shows the capital letters "NS." whereas in other plots it returns the small letter "ns" I specified.
I don't understand why this is happening and how I can fix this.

Thanks!

[schlaipferm]

I want to hide.ns in a ggplot after stat_compare_means is added, it hides all but one random capital letter NS.
install.packages("tidyverse")
library(tidyverse)
library(ggpubr)

dat2<-data.frame(R_Full_simple)
R_Full_simple

#ggplot build
dat2$grp.fac <- factor(dat2$spacer_type,levels=c("1","2","0","3","4"),
labels = c("a) non-valved \n spacer", "b) valved \n spacer",
"c) direct", "d) adapter", "e) syringe") )
dat2$grp.size <- factor(dat2$ett_size,levels=c("1","2","3","4"),
labels = c("3 mm", "5 mm", "7 mm", "9 mm"))

ett <- ggplot(dat2, outlier.shape = NA, mapping = aes(x = grp.fac, y = ett_pct, fill = grp.fac))+
facet_wrap(~grp.size, scales = "free_y")+
geom_boxplot()+geom_point()+theme_classic()
ett
#compare within group
ett+stat_compare_means(comparisons = list(c("c) direct", "a) non-valved \n spacer"),
c("a) non-valved \n spacer", "b) valved \n spacer"),
c("b) valved \n spacer", "d) adapter"), c("b) valved \n spacer",
"e) syringe"),
c("d) adapter", "e) syringe"), c("c) direct", "b) valved \n spacer"),
c("c) direct", "d) adapter"), c("c) direct", "e) syringe"),
c("a) non-valved \n spacer", "e) syringe"), c("a) non-valved \n spacer",
"d) adapter")),
label = "p.signif", hide.ns = TRUE)+
stat_compare_means(method = "anova",label.y = -2)

R_Full_simple.xlsx

I've recently come across this problem as well. The NS. is created when stat_compare_means calls geom_signif. This seems to only occur when the p-value is exactly 1 and the upper boundary of the cut values for symnum.args is 1 as well. I simply substituted a 2 for the upper value (e.g. symnum.args <- list(cutpoints = c(0, 0.05, 2), symbols = c( "*", "ns"))) and this seems to resolve the issue.

[enricoferrero]

I also came across this problem. The string NS. is displayed for p-values equal to 1, the string ns in all other (non-significant) cases. While the latter is hidden when hide.ns = TRUE, the former is not.

[kassambara]

We'll rewrite the stat_compare_means().

The following blog posts might help to add p-values onto ggplots. Consider installing the latest dev version of rstatix and ggpubr R packages.

• How to Add P-Values onto Basic GGPLOTS
• How to Add P-Values onto a Grouped GGPLOT using the GGPUBR R Package
• How to Add P-values to GGPLOT Facets