plotting on logarithmic scale (with asymptotes) and Zero in ggplot2

Question

I am plotting dose-response curves which have asymptotic tails. I would really like to include the vehicle (control) dosage in the plot showing 0

0 is generally calculated as a dosage of .0000000001 - a common practice in these plots.

I really like how the image below displays this, image was taken from a pdf on how to plot using the program: GraphPad: PRISM

Side note: I've found how to do this using basic graphics, but not using ggplot2..

A similar, but different SE question was posed regarding matlab: here

My R Code is as follows:

library(ggplot2)
library(scales)

ggplot(df, aes(x=dose,y=probability, group=model))+

  geom_ribbon(aes(ymin=Lower,ymax=Upper,x=dose,
                  fill=model, col=model,alpha=2))+

  #Axis log transformation:
  annotation_logticks(scaled = TRUE,sides="b") +
  scale_x_log10(breaks = 10^(-1:10),
                labels = trans_format("log10", math_format(10^.x)))+
  #Axes labels:
  labs(x="dosage (log scale)", y="response",size=1)

data:

df<-structure(list(dose = c(1.0000001, 1.04737100217022, 1.09698590648847,1.14895111335032, 1.20337795869652, 1.26038305255123, 1.32008852886009,1.38262230716338, 1.44811836666478, 1.51671703328309, 61.5098612858473,64.4236386159454, 67.4754441930906, 70.6718165392165, 74.0196039119089,77.525978976861, 81.1984541753771, 85.0448978198478, 89.073550951683,93.2930449978201, 97.7124202636365, 102.341145301888, 107.189137199173,112.266782823381, 117.584961077656, 123.155066208544, 128.98903221828,135.099358433491, 141.499136285126, 148.202077356965, 155.222542762819,162.575573915347, 6294.98902185499, 6593.18830115198, 6905.51354792318,7232.63392192496, 7575.25028161186, 7934.09668573241, 8309.9419660568,8703.59137460616, 9115.88830891252, 9547.71611900591, 10000),probability = c(0.000541224108467882, 0.000604351222894496,0.000674836364822755, 0.000753535950764922, 0.000841405774529429,0.000939512464066553, 0.00104904624066603, 0.00117133512397466,0.00130786074096225, 0.00146027591282991, 0.909137675722837,0.917853710549939, 0.925801889950727, 0.933037137930218,0.939612877653328, 0.945580543661554, 0.950989225589167,0.955885424480555, 0.960312903543462, 0.964312616478569,0.967922698169444, 0.971178504317829, 0.974112688435394,0.976755306362936, 0.979133940122965, 0.981273834388139,0.983198040151417, 0.984927561312444, 0.986481500855764,0.987877204102597, 0.989130397184622, 0.990255319432183,0.99999839386077, 0.999998561719608, 0.999998712035413, 0.999998846641614,0.999998967180029, 0.999999075120889, 0.99999917178077, 0.999999258338655,0.999999335850309, 0.999999405261159, 0.999999467417826),Lower = c(-0.000843143037924429, -0.000920390477371509, -0.00100418185380549,-0.0010949913193486, -0.001193314249806, -0.0012996656659587,-0.00141457798197374, -0.00153859794267898, -0.00167228258800459,-0.00181619405586472, 0.844594565230258, 0.856774199246587,0.868144364095382, 0.878738700796645, 0.888592208066634,0.897740830817022, 0.906221056637379, 0.914069534310229,0.92132272464159, 0.928016590318475, 0.934186328384788, 0.939866146372578,0.94508908115546, 0.949886858166262, 0.954289787671259, 0.958326694236121,0.962024875270718, 0.965410084524652, 0.968506536556353,0.971336928460753, 0.973922475470259, 0.976282957407176,0.999991663016519, 0.999992483777835, 0.999993224038698,0.999993891661212, 0.99999449374333, 0.999995036692715, 0.999995526293503,0.999995967766653, 0.999996365824493, 0.999996724720012,0.999997048291405), Upper = c(0.00192559125486019, 0.0021290929231605,0.002353854583451, 0.00260206322087845, 0.00287612579886485,0.0031786905940918, 0.00351267046330581, 0.0038812681906283,0.00428800406992909, 0.00473674588152455, 0.973680786215416,0.978933221853291, 0.983459415806071, 0.987335575063791,0.990633547240022, 0.993420256506086, 0.995757394540955,0.997701314650881, 0.999303082445333, 1.00060864263866, 1.0016590679541,1.00249086226308, 1.00313629571533, 1.00362375455961, 1.00397809257467,1.00422097454016, 1.00437120503212, 1.00444503810023, 1.00445646515518,1.00441747974444, 1.00433831889898, 1.00422768145719, 1.00000512470502,1.00000463966138, 1.00000420003213, 1.00000380162202, 1.00000344061673,1.00000311354906, 1.00000281726804, 1.00000254891066, 1.00000230587612,1.00000208580231, 1.00000188654425), model = c("mod3", "mod3","mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3","mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3","mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3","mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3","mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3","mod3")), .Names = c("dose", "probability", "Lower", "Upper","model"), row.names = c(1L, 2L, 3L, 4L, 5L, 6L, 7L, 8L, 9L, 10L,90L, 91L, 92L, 93L, 94L, 95L, 96L, 97L, 98L, 99L, 100L, 101L,102L, 103L, 104L, 105L, 106L, 107L, 108L, 109L, 110L, 111L, 190L,191L, 192L, 193L, 194L, 195L, 196L, 197L, 198L, 199L, 200L), class = "data.frame")

Which results in:

.

EDIT 9/1/17: below, user dww has also provided a method of doing this in r base graphics using the package plotrix

[See here](http://stackoverflow.com/questions/7194688/using-ggplot2-can-i-insert-a-break-in-the-axis) — Axeman, Jan 08 '17 at 17:25
From that question, it looked like `axis.breaks` were the solution. I am already using breaks to draw to (10^x) delineations. I made a little progress playing with the code below, but It seems ggplot2 doesn't want to draw 0...as I've already told it to draw 10^X. Perhaps it is best to not plot with that '0 break'..but now I just want to know how to do it. `scale_x_log10(breaks = c(1, 10^(2:10)), labels = trans_format("log10", math_format(10^.x)))+` — Arch, Jan 08 '17 at 17:42
`scale_x_log10(breaks = 10^(0:10), labels = c(0, math_format()(1:10)))` ? — Axeman, Jan 08 '17 at 17:54
That is very close! -All that is missing is some sort of gap between the log 10 scale and 0, in order to say this is an asymptote (as seen in first graphic) but here is the zero dose. — Arch, Jan 08 '17 at 18:07

score 9 · Accepted Answer · answered Jan 08 '17 at 18:06

9

How about something like this:

df$dose[1] <- 0.000001

df$facet <- ifelse(df$dose == min(df$dose), 1, 2)
ggplot(df, aes(x=dose,y=probability, group=model))+
  geom_point(data = subset(df, facet == 1)) +
  geom_ribbon(aes(ymin=Lower,ymax=Upper,x=dose,
                  fill=model, col=model,alpha=2))+

  #Axis log transformation:
  annotation_logticks(scaled = TRUE,sides="b") +
  scale_x_log10(breaks = c(0.000001, 10^(0:10)), 
                labels = c(0, math_format()(0:10))) +
  #Axes labels:
  labs(x="dosage (log scale)", y="response",size=1) +
  facet_grid(~facet, scales = 'free', space = 'free') +
  theme(strip.background = element_blank(),
        strip.text = element_blank())

answered Jan 08 '17 at 18:06

Axeman

32,068
8
81
94

That's it! It doesn't look as nice as I thought it would, but it is still nice to know how to do. In the case anyone is having a similar issue with this: I kept receiving an `Error... 'closure' is not subsettable`, because I forgot to rename a variable in the `df$facet <- ifelse` line. The similar question you [sent earlier](https://stackoverflow.com/questions/7194688/using-ggplot2-can-i-insert-a-break-in-the-axis) is also worth relinking I think. It briefly cautions discusses using the break. – Arch Jan 09 '17 at 12:29
1

@Arch, yes that is the downside of calling you data `df`, it is already a function and it is therefore better to avoid it and use something like `d`, `dat` or `d_f`, that would have given a more useful error. – Axeman Jan 09 '17 at 12:34

dww · Answer 2 · 2017-01-08T20:20:37.787

3

you could use plotrix for your broken axis:

library(plotrix)
plot(df$dose, df$probability, log='x', xlim=c(0.1,1e4), xaxt='n', col=NA)
axis(1, at=10^(-1:4), labels=c(0, 10^(0:4)))
axis.break(breakpos = 0.4)
polygon(c(df$dose, rev(df$dose)), c(df$Lower, rev(df$Upper)), col = rgb(0.8,0.1,0.1,0.5))
points(0.1,0, pch=19, cex=0.5)

Breaking this down:

We use xlim in the plot command to start the axis one order of magnitude below the actual data points.
We set the laels in the axis command to label the lowest point on the x axis as zero (even though it is actually at x=0.1)
We then insert an axist break between the first two labels
Finally we plot a point at (0.1, 0), which will appear to be at (0,0) because we inserted the axis break and labelled the axis to be zero here.

edited Jan 08 '17 at 20:20

answered Jan 08 '17 at 19:32

dww

30,425
5
68
111

I think OP was already aware of this option: _"Side note: I've found how to do this using basic graphics, but not using ggplot2"_. – Axeman Jan 09 '17 at 12:32
1

@Axeman ok, somehow I missed that comment in OP. Still, no harm having an answer here that shows others looking for a solution how to do it in base. More useful than a comment from OP that they know how. – dww Jan 09 '17 at 13:04
1

Yeah, I agree, I am glad you posted it. I cannot find my old r script where I had been working with r base graphics, but I believe that all I was doing in those cases was adding the term `broken=TRUE`. for example: `plot(mod_LL2.2, add=FALSE, broken=TRUE, col = "orchid")`. In certain cases iirc this worked. But anyway, I've been making the switch to ggplot2. I'll make add an edit at the bottom of the question noting your r base graphics solution. – Arch Jan 09 '17 at 18:15

plotting on logarithmic scale (with asymptotes) and Zero in ggplot2

2 Answers2