Dodged geom_errorbar with geom_bar using fill, colour, and alpha - r

Question

I am attempting to make a facet_wrap bar_graph with error bars (se) that clearly shows three different categorical variables (Treatment, Horizon, Enzyme) with one response variable (AbundChangetoAvgCtl). Below is the code for some dummy data followed by the ggplot code I have so far. The graphs I've made can be see at this link:
bargraph figures
Enzyme <- c("Arabinosides","Arabinosides","Arabinosides","Arabinosides","Arabinosides","Arabinosides","Cellulose","Cellulose","Cellulose","Cellulose","Cellulose","Cellulose","Chitin","Chitin","Chitin","Chitin","Chitin","Chitin","Lignin","Lignin","Lignin","Lignin","Lignin","Lignin")
Treatment <- c("Deep","Deep","Int","Int","Low","Low","Deep","Deep","Int","Int","Low","Low","Deep","Deep","Int","Int","Low","Low","Deep","Deep","Int","Int","Low","Low")
Horizon <- c("Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min","Org","Min")
AbundChangetoAvgCtl <- rnorm(24,mean=0,sd=1)
se <- rnorm(24, mean=0.5, sd=0.25)
notrans_noctl_enz_toCtl_summary <- data.frame(Enzyme,Treatment,Horizon,AbundChangetoAvgCtl,se)
ggplot(notrans_noctl_enz_toCtl_summary, aes(x=Horizon, y=AbundChangetoAvgCtl, fill=Horizon, alpha=Treatment)) +
geom_bar(position=position_dodge(), colour="black", stat="identity", aes(fill=Horizon)) +
geom_errorbar(aes(ymin=AbundChangetoAvgCtl-se, ymax=AbundChangetoAvgCtl+se),
width=.2,
position=position_dodge(.9)) +
scale_fill_brewer(palette = "Set1") + theme_bw() +
geom_hline(yintercept=0) +
labs(y = "Rel Gene Abundance Change / Control", x="") +
theme(axis.ticks = element_blank(),
axis.text.x = element_blank(),
strip.text.x = element_text(size=20),
plot.title = element_text(size=22, vjust=2, face="bold"),
axis.title.y = element_text(size=18),
legend.key.size = unit(.75, "in"),
legend.text = element_text(size = 15),
legend.title = element_text(size = 18)) +
facet_wrap(~Enzyme, scales="free")
(figure 1)
So this is close to what I want, however for some reason, the "alpha=Treatment" call in ggplot causes my errorbars to fade (which I don't want) as well as the bar_fill (which I do want). I've tried moving the "alpha=Treatment" to the geom_bar call, as well as adding "alpha=1" to geom_bar, but when I do that, the error bars all move to a single location and overlap (figure 2).
I initially wanted to cluster the bars within facet_wrap, but found the alpha option on this site, which seems to accomplish what I'm looking for as well. Any help would be appreciated. If there is a better way to represent all of this, those ideas are welcome as well.
Also, if there is a way to condense and clarify my legend, that would be extra bonus!
Thanks in advance for your help!
Mike

Answer 1

You need to assign Treatment to the group option in the ggplot() command and then move the alpha=Treatment option to the geom_bar() command. Then the alpha value of geom_errorbar won't be affected by the global option and will be black. Like this:
ggplot(notrans_noctl_enz_toCtl_summary, aes(x=Horizon, y=AbundChangetoAvgCtl, fill=Horizon, group = Treatment)) +
geom_bar(position=position_dodge(), colour="black", stat="identity", aes(fill=Horizon, alpha = Treatment))
Also, I would check whether setting alpha=Treatment corresponds to more transparent as being equivalent to low treatment and less transparent to high treatment. At least that would be my intuitive understanding, without any background on the research design or data.
For information about formatting legends, see here.