Plotting multiple normal curves with ggplot2 without hardcoding means and standard deviations

Question

I have a vector of means and standard deviations, and I would like to plot the densities corresponding to these means and standard deviations in the same plot using ggplot2. I used mapply and gather to solve this problem, but it's quite a lot of lines of code for something which I think should be trivial:

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

# generate data
my_data <- data.frame(mean =  c(0.032, 0.04, 0.038, 0.113, 0.105, 0.111),
                      stdev = c(0.009, 0.01, 0.01, 0.005, 0.014, 0.006), 
                      test = factor(c("Case_01", "Case_02", "Case_03", "Case_04",
                                      "Case_05", "Case_06")))

# points at which to evaluate the Gaussian densities
x <- seq(-0.05, 0.2, by = 0.001)

# build list of Gaussian density vectors based on means and standard deviations
pdfs <- mapply(dnorm, mean = my_data$mean, sd = my_data$stdev, MoreArgs = list(x = x),
               SIMPLIFY = FALSE)

# add group names
names(pdfs) <- my_data$test

# convert list to dataframe
pdfs <- do.call(cbind.data.frame, pdfs)
pdfs$x <- x

# convert dataframe to tall format
tall_df <- gather(pdfs, test, density, -x)

# build plot
p <- ggplot(tall_df, aes(color = test, x = x, y = density)) +
  geom_line() +
  geom_segment(data = my_data, aes(color = test, x = mean, y = 0, 
                                   xend = mean, yend = 100), linetype = "dashed") +
  coord_cartesian(ylim = c(-1, 100))
print(p)

This is very similar to:

Plot multiple normal curves in same plot

and as a matter of fact, the accepted answer uses mapply, so that confirms me that I'm on the right track. However, what I don't like of that answer is that it hard-codes means and standard deviations in the mapply call. This won't work in my use case, because I read the real data from disk (of course, in the MRE I skipped the data reading part for simplicity). Is it possible to simplify my code, without sacrificing readability, and without hard-coding the mean and standard deviation vectors in the mapply call?

EDIT maybe the call to mapply may be avoided by using the package mvtnorm, but I don't think that affords any real simplification here. Most of my code comes after the call to mapply.

Psidom Psidom · Accepted Answer · 2017-11-09T14:58:15

You can save some coding using purrr::pmap_df, which does the row binding automatically after constructing a data frame for each mean-stdev pair:

_{Assume my_data has the input columns in the order or mean, stdev, test and test is of character class.}

library(purrr)
tall_df2 <- pmap_df(my_data, ~ data_frame(x = x, test = ..3, density = dnorm(x, ..1, ..2)))

With data:

my_data <- data.frame(mean =  c(0.032, 0.04, 0.038, 0.113, 0.105, 0.111),
                      stdev = c(0.009, 0.01, 0.01, 0.005, 0.014, 0.006), 
                      test = c("Case_01", "Case_02", "Case_03", "Case_04", "Case_05", "Case_06"), 
                      stringsAsFactors = F)

Plot:

p <- ggplot(tall_df2, aes(color = factor(test), x = x, y = density)) + 
      geom_line() +
      geom_segment(data = my_data, aes(color = test, x = mean, y = 0, 
                                       xend = mean, yend = 100), linetype = "dashed") +
      coord_cartesian(ylim = c(-1, 100))

print(p)

gives:

Plotting multiple normal curves with ggplot2 without hardcoding means and standard deviations

1 Answers