methods to quickly turn 'tidy' data into exceedance plots (aka exceedence plots)

Question

I'm hoping to find better ways to turn a "tidy" data frame like this:

Into an "exceedance plot", or "exceedence plot" (commonly spelled both ways in water resources applications) like below, which just ranks/orders a variable along the x-axis.

Here's the lengthy way I do it now:

(sample data):

library(tidyverse)

timestep <- c("a", "b", "c", "a", "b", "c", "a", "b", "c")
var <- c("x", "x", "x", "y", "y", "y", "z", "z", "z")
taf <- c(18,1,5,23,12,67,7,30,2)
df <- data.frame(timestep, var, taf)

Build a new data frame (which I think I need?):

df_a <- df %>% filter(var == "x") %>% arrange(desc(taf))
df_b <- df %>% filter(var == "y") %>% arrange(desc(taf)) 
df_c <- df %>% filter(var == "z") %>% arrange(desc(taf)) 

df_rank <- rbind(df_a, df_b, df_c)
ts_nums <- length(unique(timestep))

taf_var_rank <- rep(seq(ts_nums),ts_nums)
taf_var_rank_xaxis <- taf_var_rank/(ts_nums+1) #standard calc for xaxis
df_rank <- data.frame(df_rank, taf_var_rank, taf_var_rank_xaxis)

Producing this, df_rank :

For my end goal of plots like these:

ggplot(df_rank, aes(x = taf_var_rank_xaxis, y = taf, color = var)) + geom_line() + 
labs(x = "probability of exceedance")

I'm pretty new to R (and programming) and I think I could build a general function, or maybe if I'm lucky there's an existing library/functions to condense this process for me? Any help is very much appreciated, as I have some long time series with many variables.

cheers, dave

Answer 1

Look at what you did, it doesn't appear you need to make those separate data.frames. You can just use mostly dplyr functions to do the same thing:

df %>% arrange(var, desc(taf)) %>% 
  group_by(var) %>% 
  mutate(taf_var_rank = row_number(),
         taf_var_rank_xaxis = taf_var_rank/(n()+1)) %>% 
  ggplot(aes(x = taf_var_rank_xaxis, y = taf, color = var)) + 
    geom_line() + 
    labs(x = "probability of exceedance")