Plotting in Pumas

How-to
1

Unfortunately I cannot share data. This is a box plot with observed data points scattered with jitter.

isMD.labs <- c("SD", "MD")
names(isMD.labs) <- c("0", "1")

ggplot(tab_all, aes(x = as.factor(dose), y = cfb, color=as.factor(dose)))+
  geom_boxplot(outlier.shape=NA)+
  geom_jitter(shape=16,size=0.4, position=position_jitter(0.2))+
  #scale_x_continuous(breaks = seq(0,20,6),
  #                     limit = c(0,21))+
  scale_y_continuous(breaks = seq(-25,50,25),
                       limit = c(-25,50))+
  labs(x = "Dose (mg)", 
       y = "CFB")+
  theme_bw(base_size = 16)+
  theme(legend.position = "none") +
  facet_wrap(~isMD, labeller = labeller(isMD = isMD.labs))
2

You can do the same with AlgebraOfGraphics.jl which is a package inspired by ggplot2:

plt = data(tab_all) *
  mapping(
    :dose => nonnumeric,
    :cfb;
    color=:dose => nonnumeric,
    layout=:isMD => nonnumeric,
) *
  visual(BoxPlot)

draw(plt;
     axis=(; yticks=-25:50:25, limits=(nothing, nothing, -25, 50)),
     figure=(; fontsize=16),
     legend=(; position=:top, titleposition=:left),
)

We also have tons of publication-ready easy-to-use visualization functions for most of the pharmacometrics workflows. Don’t forget to check Plotting · Pumas.

3

Thank you @storopoli Three questions:

  1. Do we need to load AlgebraOfGraphics.jl ?
  2. I am assuming data() is a plots function; and tab_all is my dataframe? I get an error that ‘data’ does not exist.
  3. Can we pass pumas object into data()? For eg. can I pass inspect(myfit) or do I need to pass ‘DataFrame(inspect(myfit))’.
    J
4

Yes in the same way you need to do library(ggplot) to call the ggplot() function in the original code. You need to do:

using AlgebraOfGraphics
using CairoMakie # backend

This is fixed with the using AlgebraOfGraphics above

Almost all Pumas’ outputs can be easily converted to DataFrames with DataFrame(pumas_output).
So you can definitely do that with:

my_inspect = inspect(my_fit)
DataFrame(my_inspect)