ggplot2 - Getting multiple error bars in bar graph (R)

Question

I am trying to plot a bar graph in R with 4 independent variables - time(t1,t2), group(1,2,3,4,5), distance(far and near) and cue(valid and invalid) with RT as the dependent variable. For the same, I have used the following code

ggplot(b, aes(x=cue, y=RT, fill = cue))+
  geom_bar(stat="identity", position = position_dodge(),  width = .9)+
  facet_grid(group~time,  space="free_x") +
  geom_errorbar(aes(ymin= RT-se, ymax = RT+se), width = 0.2, color = "BLACK", position=position_dodge())+
  coord_cartesian(ylim = c(200,1500))+theme(legend.title = element_blank())

When running the codes in R, I am getting the following graph

Plot here - bar plot

Is it possible to rearrange cue (valid/invalid as well as distance (near/far) in a descending manner (both to be done together).

The error bars seem to be off centre, how do I fix it? Also, can I statistically compare two items (for example, comparing valid and invalid under group 1, time1) and denote them in the graph?

---

The data set looks something like this for each participant:

participant	cue	distance	RT	time	group
P1	valid	far	1461	T1	4
P1	invalid	near	1416	T1	4
P1	invalid	near	1409	T1	4
P1	invalid	far	1351	T1	4

question from:https://stackoverflow.com/questions/65913384/getting-multiple-error-bars-in-bar-graph-r

Answer 1

#--------------------------

Answer after question edits.

Error bar alignment is done by the call to position_dodge. Invalid/valid reordering is a call to fct_rev. Statistical comparison depends on what you actually want to show, and then try to figure out how you want to show it.

library(tidyverse)
b <- tribble(
  ~participant, ~cue, ~distance, ~RT, ~time, ~group,
  "P1", "valid",    "far",  1461,   "T1",   4,
  "P1", "invalid",  "near", 1416,   "T1",   4,
  "P1", "invalid",  "near", 1409,   "T1",   4,
  "P1", "invalid",  "far",  1351,   "T1",   4,
  "P1", "invalid",  "far",  1391,   "T1",   4,
  "P1", "invalid",  "far",  1365,   "T1",   4,
  "P1", "invalid",  "far",  1385,   "T1",   4,
  "P1", "invalid",  "near", 1465,   "T1",   4,
  "P1", "valid",    "near", 1451,   "T1",   4,
  "P1", "valid",    "near", 1397,   "T1",   4,
  "P1", "valid",    "far",  1466,   "T1",   4,
  "P1", "invalid",  "far",  1411,   "T1",   4,
  "P1", "invalid",  "near", 1439,   "T1",   4,
  "P1", "valid",    "far",  1328,   "T1",   4,
  "P1", "valid",    "far",  1437,   "T1",   4,
  "P1", "valid",    "far",  1376,   "T1",   4,
  "P1", "invalid",  "far",  1364,   "T1",   4,
  "P1", "invalid",  "near", 1451,   "T1",   4,
  "P1", "valid",    "far",  1461,   "T1",   4,
  "P1", "invalid",  "far",  1441,   "T1",   4,
  "P1", "valid",    "near", 1491,   "T1",   4,
  "P1", "valid",    "near", 1385,   "T1",   4,
  "P1", "valid",    "near", 1553,   "T1",   4,
  "P1", "invalid",  "far",  1484,   "T1",   4,
  "P1", "valid",    "far",  1449,   "T1",   4,
  "P1", "invalid",  "near", 1361,   "T1",   4,
  "P1", "invalid",  "near", 1399,   "T1",   4,
  "P1", "invalid",  "near", 1389,   "T1",   4,
  "P1", "valid",    "near", 1378,   "T1",   4,
  "P1", "valid",    "near", 1365,   "T1",   4,
  "P1", "valid",    "far",  1465,   "T1",   4,
  "P1", "valid",    "near", 1333,   "T1",   4,
  "P1", "valid",    "near", 1340,   "T1",   4,
  "P1", "invalid",  "far",  1347,   "T1",   4,
  "P1", "valid",    "far",  1375,   "T1",   4,
  "P1", "valid",    "near",  390,   "T2",   4,
  "P1", "invalid",  "far",   394,   "T2",   4,
  "P1", "invalid",  "near",  374,   "T2",   4,
  "P1", "valid",    "far",   363,   "T2",   4,
  "P1", "valid",    "near",  342,   "T2",   4,
  "P1", "invalid",  "far",   421,   "T2",   4,
  "P1", "invalid",  "near",  398,   "T2",   4,
  "P1", "invalid",  "near",  419,   "T2",   4
)


b %>% 
  group_by(participant, cue, distance, time, group) %>% 
  summarise(RT_mean = mean(RT), 
            RT_sd = sd(RT)) %>% 
  filter(participant == "P1") %>% #not strictly necessary in this instance, but will be in general.
  mutate(cue = fct_rev(cue)) %>% 
ggplot(aes(x=cue, y=RT_mean, fill = distance))+
  geom_bar(stat="identity", position = position_dodge(),  width = .9)+
  facet_grid(group~time,  space="free_x") +
  geom_errorbar(aes(ymin= RT_mean - RT_sd, ymax = RT_mean + RT_sd), 
                width = 0.2, color = "BLACK", 
                position=position_dodge(0.9))+ #the 0.9 here should the same value as the width in geom_bar
                                              #  to keep the error bar centred.
  coord_cartesian(ylim = c(200,1500))+theme(legend.title = element_blank())
#> `summarise()` regrouping output by 'participant', 'cue', 'distance', 'time' (override with `.groups` argument)

^{Created on 2021-01-27 by the reprex package (v0.3.0)}

#--------------------------

Original answer

I've made a dummy dataset, and think you may have a data quality issue. See that the first two rows of b have the same cue, time, and group, but different RT. Can you post your original data?

The ordering of "valid"/"invalid" can be reversed using the forcats package, as in my second example.

library(tidyverse)
cue <- c("invalid",   "invalid","valid","invalid","valid","invalid","valid","invalid","valid")
time <- c("T1",           "T1","T1","T2","T2","T1","T1","T2","T2")
group <- c(1,                1,1,1,1,2,2,2,2)
RT <- c(1000,             1200,1300,400,500,700,800,300,400)
ci <- c(50,                100,100,100,100,50,50,50,50)

b <- tibble(cue,time,group,RT,ci)

b
#> # A tibble: 9 x 5
#>   cue     time  group    RT    ci
#>   <chr>   <chr> <dbl> <dbl> <dbl>
#> 1 invalid T1        1  1000    50
#> 2 invalid T1        1  1200   100
#> 3 valid   T1        1  1300   100
#> 4 invalid T2        1   400   100
#> 5 valid   T2        1   500   100
#> 6 invalid T1        2   700    50
#> 7 valid   T1        2   800    50
#> 8 invalid T2        2   300    50
#> 9 valid   T2        2   400    50

ggplot(b,aes(x=cue, y=RT, fill = cue))+
  geom_bar(stat="identity", position = position_dodge(),  width = .9)+
  facet_grid(group~time,  space="free_x") +
  geom_errorbar(aes(ymin= RT - ci, ymax = RT + ci), width = 0.2, color = "BLACK", position=position_dodge())+
  coord_cartesian(ylim = c(200,1500))+theme(legend.title = element_blank())

#reverse the order of the "invalid"/"valid"
b %>% 
  mutate(cue = fct_rev(cue)) %>% 
ggplot(aes(x=cue, y=RT, fill = cue))+
  geom_bar(stat="identity", position = position_dodge(),  width = .9)+
  facet_grid(group~time,  space="free_x") +
  geom_errorbar(aes(ymin= RT - ci, ymax = RT + ci), width = 0.2, color = "BLACK", position=position_dodge())+
  coord_cartesian(ylim = c(200,1500))+theme(legend.title = element_blank())

^{Created on 2021-01-27 by the reprex package (v0.3.0)}