Labels on wrong dodged columns in geom_col

I want to create a simple barplot of my data frame:

> dput(corr)
structure(list(`sample length` = structure(c(2L, 2L, 2L, 2L, 
2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 2L, 1L, 1L, 1L, 1L, 1L, 1L, 1L, 
1L, 1L, 1L, 1L), .Label = c("3s", "10s"), class = "factor"), 
    feature = structure(c(1L, 1L, 5L, 5L, 2L, 5L, 6L, 5L, 5L, 
    4L, 1L, 1L, 1L, 1L, 1L, 2L, 5L, 5L, 3L, 4L, 1L, 1L, 1L, 1L
    ), .Label = c("f0", "f1", "f2", "f3", "f2 prime", "f2-f1"
    ), class = "factor"), measure = c("meanf0 longterm", "meanf0 longterm st", 
    "f2' Fant", "f2' Carlson", "F1meanERB", "F2meanERB", "f2-f1 ERB", 
    "f2' Fant", "f2' Carlson", "F3meanERB", "meanf0 3secs", "meanf0 3secs st", 
    "meanf0 10secs", "meanf0 longterm", "meanf0 longterm st", 
    "F1meanERB", "f2' Fant", "f2' Carlson", "F2meanERB", "F3meanERB", 
    "meanf0 longterm", "meanf0 longterm st", "meanf0 3secs", 
    "meanf0 3s st"), score = c(0.574361009949897, 0.592472685498182, 
    0.597453479834514, 0.529641256460457, 0.585994252821649, 
    0.618734735308094, 0.517715270144259, 0.523916918327387, 
    0.616237363007349, 0.732926257362305, 0.649505366093518, 
    0.626628120773466, 0.522527636952945, 0.53968850323167, 0.548664887822775, 
    0.648294358978928, 0.650806695307235, 0.696797693503567, 
    0.621298393945597, 0.57140950987443, 0.606634531002859, 0.597064217305556, 
    0.582534743353082, 0.572808145210493), dimension = structure(c(1L, 
    1L, 1L, 1L, 1L, 1L, 3L, 3L, 3L, 3L, 4L, 4L, 4L, 1L, 1L, 1L, 
    2L, 2L, 2L, 2L, 4L, 4L, 4L, 4L), .Label = c("1", "2", "3", 
    "4"), class = "factor")), row.names = c(NA, -24L), class = c("tbl_df", 
"tbl", "data.frame"))

I have tried the following code:

ggplot(data=corr, aes(x=factor(dimension), y=score)) + 
  geom_col(aes(fill=feature),position=position_dodge2(width=1,preserve='single')) +
  facet_grid(~`sample length`, scales='free_x',space='free_x') +
  labs(x="Dimension", y="Correlation Coefficient (Abs. value)") +
  geom_text(aes(label=measure),position=position_dodge2(width=0.9, preserve='single'), angle=90,
            size=4,hjust=2.5,color='white')

Giving the following barplot:

However, the labels for 'measure' are being incorrectly assigned to the columns. E.g. for 3s facet plot, under 'dimension 2', the two light blue bars should be labelled as 'f2' Carlson' and 'f2' Fant' but they have been swapped with the other two labels.

I think the levels must be wrong, but I don't understand how!

Any help much appreciated, ta

The problem of switching labels comes from geom_text() not knowing how the information should be split for the purposes of dodging. The solution is to supply a group= aesthetic to geom_text() that matches the fill= aesthetic specified for geom_col().

In the case of geom_col(), you specify aes(fill=feature). The height of the different columns is therefore grouped automatically by corr$feature. You can supply a group= aesthetic as well, but it's unnecessary and the dodging will happen as you expect.

In the case of geom_text(), there is no obvious way to group the data. When you do not specify a group= aesthetic, ggplot2 chooses one of the columns (in this case, the first column number) for grouping. For dodging to work here, you need to specify how the label information is grouped. If you don't have a specific legend-associated aesthetic to choose here, you can use the group= aesthetic to specify group=feature. This let's ggplot2 know that the text labels should be sorted and dodged by grouping according to this column in the data:

ggplot(data=corr, aes(x=factor(dimension), y=score)) +
  geom_col(aes(fill=feature),position=position_dodge2(width=1,preserve='single')) +
  facet_grid(~`sample length`, scales='free_x',space='free_x') +
  labs(x="Dimension", y="Correlation Coefficient (Abs. value)") +
  geom_text(aes(label=measure, group=feature),position=position_dodge2(width=0.9, preserve='single'), angle=90,
            size=4,hjust=2.5,color='white')

As a side note, you don't have to specify the group= aesthetic if you assign a color-based aesthetic (or one that would result in a legend). If we set color=feature with geom_text(), it works without group=. To see the labels, you need to set the alpha for the columns a bit lower, but this should illustrate the point well:

ggplot(data=corr, aes(x=factor(dimension), y=score)) + 
  geom_col(aes(fill=feature),position=position_dodge2(width=1,preserve='single'), alpha=0.2) +
  facet_grid(~`sample length`, scales='free_x',space='free_x') +
  labs(x="Dimension", y="Correlation Coefficient (Abs. value)") +
  geom_text(aes(label=measure, color=feature),position=position_dodge2(width=0.9, preserve='single'), angle=90,
            size=4,hjust=2.5)