tip_length not used when geom_signif() with stat = "identity" about ggsignif HOT 4 CLOSED

I can confirm that having control over the tip_length when using stat = 'identity' would be very useful
Thanks for looking into it!

from ggsignif.

Hi thanks for raising the issue. You are right that this is confusing that tip_length has no effect if stat = "identity", but this is actually the intended behavior.

If stat="signif" I am internally computing the correct starting and end points for the lines which are subsequently passed to the geom_signif(stat="identity", ...).

So to add tips if stat = "identity" you will need to do something like this:

dat <- data.frame(Group = c("S1", "S1", "S2", "S2"),
                  Sub   = c("A", "B", "A", "B"),
                  Value = c(3,5,7,8))

ggplot(dat, aes(Group, Value)) +
      geom_bar(aes(fill = Sub), stat="identity", position="dodge", width=.5) +
      geom_signif(stat="identity",
                  data=data.frame(x=c(0.875, 0.875, 1.125), xend=c(0.875, 1.125, 1.125),
                                  y=c(5.3, 5.8, 5.8), yend=c(5.8, 5.8, 5.3),
                                  annotation=c("***")),
                  aes(x=x,xend=xend, y=y, yend=yend, annotation=annotation))
  

I understand that this might seem tedious, so I am considering adding another stat (maybe stat="signifmanual"), which simplifies this but this would be a little bit more work.

from ggsignif.

Indeed this makes sense. I hadn't realized the length of y and yend wasn't necessarily the same as the data in your code here https://github.com/Artjom-Metro/ggsignif/blob/master/R/significance_annotation.R#L71

If it's just a matter of passing a tip extremity as 1st value e.g. y = c(y_lower_tip, y_bar, y_bar) then it's not as bad as I thought: I feared the whole layer had to be altered, and since I'm not familiar with how they are coded, i was expectign something complicated.

This might warrant a mention in the vignette though, maybe replace the identity example with the one you just posted here including a comment on the length of y/yend vectors.

Thanks for the input though!

from ggsignif.

Hi, I have just released a new version 0.3.0 that simplifies the definition of custom locations, by adding a xmin and xmax parameter to the function.

dat <- data.frame(Group = c("S1", "S1", "S2", "S2"),
                  Sub   = c("A", "B", "A", "B"),
                  Value = c(3,5,7,8))

ggplot(dat, aes(Group, Value)) +
  geom_bar(aes(fill = Sub), stat="identity", position="dodge", width=.5) +
  geom_signif(annotation="***", 
              y_position=5.8, xmin=0.875, xmax=1.125, tip_length = 0.01)

As you see, it is no longer necessary to use stat="identity" and also the tip_length parameter works.

from ggsignif.