How do you recenter y-axis at 0.01 using ggplot2-CodePudding

I am trying to make a grouped bar plot comparing the concentration of contaminants (in ug/g) from 2 different locations site comparison figure, with the y axis in log scale. I have some values that are below one but greater than zero. The y axis is centering my bars at y=1, making some of the bars look negative. Is there to make my bars start at 0.01 instead of 1?

I tried coord_cartesian( ylim=c(0.01,100), expand = FALSE ) but that didn't do it. I also tried to use coor_trans(y='log10') to log transform y axis instead of scale_y_continuous(trans='log10') but was getting the error message "Transformation introduced infinite values in y-axis" even though I have no zero values.

Any help would be much appreciated, Thank you.

my code is below:

malcomp2 %>% 
  ggplot(aes(x= contam, y= ug_values, fill= Location)) 
  geom_col(data= malcomp2,
           mapping = aes(x= contam, y= ug_values, fill = Location),
           position = position_dodge(.9), #makes the bars grouped
           stat_identity(),
           colour = "black", #adds black lines around bars
           width = 0.8,
           size = 0.3) 
  ylab(expression(Concentration~(mu*g/g)))  #adds mu character to y axis
  coord_cartesian( ylim=c(0.01,100), expand = FALSE )    #force bars to start at 0
  theme_classic()  #get rid of grey grid
  scale_y_continuous(trans='log10',   #change to log scale
                     labels = c(0.01,0.1,1,10,100))  #change axis to not sci notation
  annotation_logticks(sides = 'l')     #add log ticks to y axis only
  scale_x_discrete(name = NULL, #no x axis label
                   limits = c('Dieldrin','Mirex','PBDEs','CHLDs','DDTs','PCBtri','PCBquad',
                            'PCBhept'),   #changes order of x axis
                   labels = c('Dieldrin','Mirex','PBDEs','CHLDs','DDTs','PCB 3','PCB 4-6','PCB7 '))    
  scale_fill_manual("Location",      #rename legend
                    values = c('turquoise2','gold'),  #change colors
                    labels = c( 'St.Andrew Bay', 'Sapelo'))   #change names on legend
  theme(legend.title = NULL,
        legend.key.size = unit(15, "pt"),
        legend.position =   c(0.10,0.95))   #places legend in upper left corner

CodePudding user response：

One hack would be to just scale your data so baseline is at 1. People have asked this question before on SO, and it seems like a more satisfactory approach might be to use geom_rect instead, like here: