0

I am trying to plot 3 models with group=model, but only the last model in the data.frame is being plotted. I've done a fair amount of searching online but haven't been able to figure out what the problem is.

The data.frame has 3 columns: dose(x axis), p1(y axis), and model(mod1,mod2,mod3)

library(ggplot2)

ggplot(df, aes(x=dose,y=p1, group=model))+
   geom_line(aes(color=model))+
   geom_step(aes(x=dose,y=p1,group=model, col=model))+
   coord_trans(x = 'log10', limx = c(0.01,200), limy=c(0.01,0.05)) +

#Assigning colors to mod1,mod2,mod3
   values = c("mod1" = "darkgreen", "mod2" = "blue", "mod3" = "red"))

#Log Aesthetics, cut for brevity of code, Help with log ticks [enter image description here]

-It isn't an issue of mod1 and mod2 being plotted outside of the frame.

If it is of any relevance: The data I have included below (which I created the graphic above with) only includes a few points from the lower bound of 3 models - The actual data set contains 8 models, each with 100 points. 

df <- structure(list(dose = c(0.5, 0.565608284362011, 0.639825462677876, 
  0.723781164472726, 0.818753245381915, 0.5, 0.565608284362011, 
  0.639825462677876, 0.723781164472726, 0.818753245381915, 0.926187236872587, 
  0.5, 0.565608284362011, 0.639825462677876, 0.723781164472726, 
  0.818753245381915, 0.926187236872587, 1.04771834809099, 1.18519635471669, 
  1.34071375364684, 1.51663761204148, 1.71564559549136), p1 = c(0.0103075076812739, 
  0.0116952370538794, 0.0132672958208565, 0.0150474513444454, 0.017062331184752, 
  0.0103075076812739, 0.0116952370538794, 0.0132672958208565, 0.0150474513444454, 
  0.017062331184752, 0.0193417088954045, 0.0103075076812739, 0.0116952370538794, 
  0.0132672958208565, 0.0150474513444454, 0.017062331184752, 0.0193417088954045, 
  0.0219188012260592, 0.0248305714535104, 0.0281180313564242, 0.0318265316115288, 
  0.036006027058739), model = c("mod1", "mod1", "mod1", "mod1", 
  "mod1", "mod2", "mod2", "mod2", "mod2", "mod2", "mod2", "mod3", 
  "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", "mod3", 
  "mod3", "mod3")), .Names = c("dose", "p1", "model"), row.names = c(1L, 
  2L, 3L, 4L, 5L, 101L, 102L, 103L, 104L, 105L, 106L, 201L, 202L, 
  203L, 204L, 205L, 206L, 207L, 208L, 209L, 210L, 211L), class = "data.frame")
cuttlefish44
  • 6,586
  • 2
  • 17
  • 34
Arch
  • 192
  • 2
  • 16

1 Answers1

1

From what I understand the problem is that the points are over each other, meaning that you can only see the more complete line (mod3) and the other ones are underneath.

White some basic data wrangling we can get rid of the duplicates and get something like this

library(ggplot2)
require(tidyr)
require(dplyr)

 df1 <- df %>% spread(model,p1) %>% 
  mutate(mod3=ifelse(is.na(mod2*mod3)==FALSE,NA,mod3),mod2=ifelse(is.na(mod2*mod1)==FALSE,NA,mod2)) 
%>% gather(model,p1,mod1:mod3) %>% filter(is.na(p1)==FALSE) %>% arrange(p1) 

df1 <- rbind(df1,df1[which(df1$p1==max(df1[df1$model=="mod1",]$p1)),])
df1 <- rbind(df1,df1[which(df1$p1==min(df1[df1$model=="mod3",]$p1)),])
df1[nrow(df1),2] <- "mod2"
df1[nrow(df1)-1,2] <- "mod2"


ggplot(df1, aes(x=dose,y=p1, group=model))+
geom_line(aes(color=model))+
geom_step(aes(x=dose,y=p1,group=model, col=model))+
coord_trans(x = 'log10', limx = c(0.01,200), limy=c(0.01,0.05))

Outputs the following image:

enter image description here

erc
  • 10,113
  • 11
  • 57
  • 88
eclark
  • 819
  • 7
  • 16
  • 2
    How about using facetting to show three models side by side? Based on what phenomenon one would like to show, you can flank it or dodge it up/down. – Roman Luštrik Oct 12 '16 at 05:40
  • It turned out to be an issue where I was neglecting to change a variable name - I was renaming each predicted model `pm1<- predict(mod1, newdata=newdata1,interval="confidence")` – Arch Oct 12 '16 at 20:42
  • ... where I was renaming each extracted dataset 'pm1'. But you've showed me some advanced tricks, thanks! @RomanLuštrik Just for learning more ggplot2, How do you mean "show three models side by side"? using something like `multiplot` ? – Arch Oct 12 '16 at 20:44
  • 1
    @RomanLuštrik is refering to this: http://www.cookbook-r.com/Graphs/Facets_(ggplot2)/ . It's also a good way to plot different groups side by side – eclark Oct 12 '16 at 20:57