How to add manually pvalue on geom_barplot with group? - r

Question

I have this dataframe
bat=rep(c("A", "B", "C", "D"),3)
loc=c(rep("ab",4), rep("te",4), rep("po", 4))
value=c(17,8,14,20,2,9,11,18,5,17,7,14)
tot=rep(c(30,45,36,58),3)
say=rep()
dt=data_frame(bat, loc, value, tot) %>% mutate(pct=value/tot*100) %>% mutate(se=sqrt(((pct*(100-pct))/(tot)))) %>% mutate(value2=tot-value)
> dt
# A tibble: 12 x 7
bat loc value tot pct se value2
<chr> <chr> <dbl> <dbl> <dbl> <dbl> <dbl>
1 A ab 17 30 56.7 9.05 13
2 B ab 8 45 17.8 5.70 37
3 C ab 14 36 38.9 8.12 22
4 D ab 20 58 34.5 6.24 38
5 A te 2 30 6.67 4.55 28
6 B te 9 45 20 5.96 36
7 C te 11 36 30.6 7.68 25
8 D te 18 58 31.0 6.07 40
9 A po 5 30 16.7 6.80 25
10 B po 17 45 37.8 7.23 28
11 C po 7 36 19.4 6.60 29
12 D po 14 58 24.1 5.62 44
I made fisher test between each pair of bat for each loc with this following code.
dt_ab=dt %>% filter(loc=="ab")
dt_po=dt %>% filter(loc=="po")
dt_te=dt %>% filter(loc=="te")
# ab
idx = t(combn(seq_along(dt_ab$bat),2))
res_ab = lapply(1:nrow(idx),function(i){
test = fisher.test(dt_ab[idx[i,],c("value","value2")])
data.frame(
group1 = dt_ab$bat[idx[i,1]],
group2 = dt_ab$bat[idx[i,2]],
loc=dt_ab$loc[idx[i,1]],
odds_ratio = as.numeric(test$estimate),
p = as.numeric(test$p.value)
)
})
res_ab = do.call(rbind,res_ab)
res_ab
# po
idx = t(combn(seq_along(dt_po$bat),2))
res_po = lapply(1:nrow(idx),function(i){
test = fisher.test(dt_po[idx[i,],c("value","value2")])
data.frame(
group1 = dt_po$bat[idx[i,1]],
group2 = dt_po$bat[idx[i,2]],
loc=dt_po$loc[idx[i,1]],
odds_ratio = as.numeric(test$estimate),
p = as.numeric(test$p.value)
)
})
res_po = do.call(rbind,res_po)
res_po
# te
idx = t(combn(seq_along(dt_te$bat),2))
res_te = lapply(1:nrow(idx),function(i){
test = fisher.test(dt_te[idx[i,],c("value","value2")])
data.frame(
group1 = dt_te$bat[idx[i,1]],
group2 = dt_te$bat[idx[i,2]],
loc=dt_te$loc[idx[i,1]],
odds_ratio = as.numeric(test$estimate),
p = as.numeric(test$p.value)
)
})
res_te= do.call(rbind,res_te)
res_te
res=NULL
res=rbind(res_ab, res_po, res_te)
res
group1 group2 loc odds_ratio p
1 A B ab 5.8817685 0.0009288055
2 A C ab 2.0321378 0.2157927844
3 A D ab 2.4578341 0.0678462682
4 B C ab 0.3445393 0.0451456088
5 B D ab 0.4143084 0.0750566107
6 C D ab 1.2066231 0.6665182345
7 A B po 0.3341536 0.0700086821
8 A C po 0.8309230 1.0000000000
9 A D po 0.6317547 0.5859688210
10 B C po 2.4870606 0.0896596469
11 B D po 1.8959538 0.1934540389
12 C D po 0.7608147 0.7994419705
13 A B te 0.2899040 0.1823689517
14 A C te 0.1664441 0.0270616991
15 A D te 0.1614577 0.0141330017
16 B C te 0.5722244 0.3084931936
17 B D te 0.5586957 0.2609751992
18 C D te 0.9780082 1.0000000000
Now I would like to add manually the pvalue and the odd ratio for each pair on the following barplot
I saw we can use the stat_pvalue_manual function to do that, but I have some difficulties to use it. Please, someone can have look on my code and tell me what is wrong. Below is the code and the plot I obtained.
bp_tab_pct<- ggplot(dt, aes(x=loc, y=pct, fill=cat))+
geom_bar(stat = "identity",position="dodge") +
scale_fill_manual(values = c("A"= "#5977FF", "B"="#FF7F50", "C"= "#00CED1", "D" = "#FFFF33"))+
theme(plot.title = element_text(color="black", size=15, face="bold.italic")) +
labs(fill = "") +
ylim(c(0,100))+
ggtitle(label = "")+
geom_text(aes(y = pct+se+2, label = paste(round(pct,2),"%", "\n","(",value,"/",tot,")")), color = "black",size=5, position = position_dodge(width = 0.9))+
theme(plot.title = element_text(color="black", size=15, face="bold.italic"))+
theme(axis.text.x = element_text(size=16), axis.text.y = element_text(size=14))+
theme(axis.title.y = element_text(size=16), axis.title.x = element_blank(),) +
geom_errorbar(aes(ymin=pct-se, ymax=pct+se), width=.2,
position=position_dodge(.9)) +
stat_pvalue_manual(res, y.position = 35, step.increase = 0.1,
label = "p", inherit.aes = FALSE)
enter image description here
I also tried with facet insted of group but I still don't have the plot I want.
It seems that the stat_pvalue_manual function considere only the 10 first lines and I don't know why.
bp_tab_pct<- ggplot(dt, aes(x=bat, y=pct, fill=bat))+
facet_wrap(~loc) +
geom_bar(stat = "identity",position="dodge") +
scale_fill_manual(values = c("A"= "#5977FF", "B"="#FF7F50", "C"= "#00CED1", "D" = "#FFFF33"))+
theme(plot.title = element_text(color="black", size=15, face="bold.italic")) +
labs(fill = "") +
ylim(c(0,100))+
ggtitle(label = "")+
geom_text(aes(y = pct+se+2, label = paste(round(pct,2),"%", "\n","(",value,"/",tot,")")), color = "black",size=5, position = position_dodge(width = 0.9))+
theme(plot.title = element_text(color="black", size=15, face="bold.italic"))+
theme(axis.text.x = element_text(size=16), axis.text.y = element_text(size=14))+
theme(axis.title.y = element_text(size=16), axis.title.x = element_blank(),) +
geom_errorbar(aes(ymin=pct-se, ymax=pct+se), width=.2,
position=position_dodge(.9)) +
stat_pvalue_manual(res, y.position = 35, step.increase = 0.1,
label = "p", inherit.aes = FALSE)
enter image description here

Answer 1

I understood the problem. It was du to the ylim masking the highest values on the y axis. Now I would like to be able to adjust for the y position of each pvalue as I did for the pct indicated up to the bar but I will post another question for that
enter image description here