Use ggplot2 to plot multiple curves

Question

Say I have a data frame coefs where each row contains model coefficients for a curve.

coefs <- structure(list(a1 = c(1.22228259789383, 1.2064168157394, 1.09555089661994, 0.943947433470916, 0.883490658557721, 0.46125552320107), d = c(0.385227755933488, 0.457271644919152, 0.340063262461958, 0.305629949064525, 0.42459163183877, 0.425710112988664), g = c(0, 0, 0, 0, 0, 0), u = c(1, 1, 1, 1, 1, 1)), .Names = c("a1", "d", "g", "u"), class = c("tbl_df", "tbl", "data.frame"), row.names = c(NA, -6L))

I'd like to use each row of the data frame to add a new curve to the plot based a defined function: (you may recognize it as a 2PL item response model)

TWOPL <- function(x,a1,b) {
  1 / (1 + exp(-a1*(x-(b))))
}

Based on this and this question I tried the following ggplot command but get the error that computation failed:

library(ggplot2)
p <- ggplot(coefs, aes(x = 0))
p + stat_function(fun = TWOPL) + xlim(-5,5)

I know that I need a way to give the various coefficients to the function. As a test, I tried the function with a fixed parameters to create 1 curve and it works, for example:

#1 curve based on fixed parameters
TWOPL_copy <- function(x) {
  1 / (1 + exp(-1.22*(x-(.385))))
}

p <- ggplot(data.frame(x = 0), aes(x = 0))
p + stat_function(fun = TWOPL_copy) + xlim(-5,5)

I'm wondering how I might send each row of the data frame to ggplot. An ideal next step would be to differentiate the colors of each line somehow.

I just came across this question (stackoverflow.com/questions/21686638/…) and a possible approach is to make a temporary data frame that extrapolates the y values, then pass that to ggplot2. — Andrew Jackson
– Andrew Jackson, Commented Mar 31, 2017 at 4:53

alistaire · Accepted Answer · 2018-04-10 15:23:18Z

While you could call stat_function for each set of parameters or with some pain call it programmatically, it's simpler to do the calculations yourself:

library(tidyverse)

coefs %>% 
    mutate(curve = letters[row_number()]) %>%    # add curve name
    crossing(x = seq(-5, 5, .1)) %>%    # repeat each row for every occurence of x
    mutate(y = TWOPL(x, a1, d)) %>%    # compute y values
    ggplot(aes(x, y, color = curve)) + 
    geom_line()

The simplest way to create the curves programmatically is to add a list of stat_function calls to the plot. All aesthetics have to be iterated across, including color. An x aesthetic must be supplied, but if you set xlim, it doesn't matter what it is.

curves <- coefs %>% 
    mutate(curve = letters[row_number()]) %>% 
    pmap(function(...){
             dots <- data_frame(...)
             stat_function(data = dots, aes(0, color = curve), 
                           fun = function(x) TWOPL(x, dots$a1, dots$d), 
                           xlim = c(-5, 5))
    })

ggplot() + curves

Collectives™ on Stack Overflow

Use ggplot2 to plot multiple curves

1 Answer 1

Comments

Linked

Hot Network Questions

Collectives™ on Stack Overflow

1 Answer 1

Comments

Linked

Related