Creating stimuli in R with ggplot

1
votes

I am trying to generate my own stimuli for an experiment using R. Below is the code that creates my (x,y) coordinates using the rnorm() with different a sample size of 100, different means and sd. I also create another variable to represent the size of the circles, which are determined by the runif().

dt <- data.frame(x = NA,
         y = NA,
         size = NA,
         M = NA, 
         sd = NA,
         col = NA,
         iter = NA)
sa<-0


mySD<-c(5, 15)
myMeans<-c(35, 45)
colors<-c("Blues", "Reds") 

for(i in 1:10){
    for(s in mySD){
      for(m in myMeans){

        x = abs(rnorm(n=1, mean=m, sd=s))
        y = abs(rnorm(n=1, mean=m, sd=s))

        size = runif(1, 1, 25) #select a random x speed between [25,35]

        sa<-sa+1
        dt[sa,] <- NA
        dt$x[sa]<-x
        dt$y[sa]<-y
        dt$M[sa]<-m
        dt$sd[sa]<-s
        dt$size[sa]<-size
        dt$iter[sa]<-i
      }
    }
  }
}

Next, I want to use ggplot(dt, aes(x, y, size=size) to plot. I want to randomly select 4 (x,y) values to plot for one graph, then 8 for another, then 16 for another, etc. Basically, I want to plot different graphs with a different number of data points. For example, some graphs that you would see would have 4 data points that vary by size and color, others would have 32 data points that vary in size and color. I m not sure how to select a set of unique data points from the data frame that I created. Any help would be great. I'm pretty new to R.

2
rggplot2graph
Why is n = 100 in x = abs(rnorm(n=100, mean=m, sd=s)) if you are assigning only one value at a time in dt$x[sa]<-x? – Rui Barradas
Thanks for your reply. I had thought rnorm() selected one random value from the specified sample size. I guess that is only runif(). I'll make that change now, but that still doesn't get at what I'm trying to ultimately do. – sara connor
Do you want each group to not contain points from any other group? – Paul
Thanks for the question, Paul. Yes, I want them to be unique for each group (trial). Would that mean that I should use: x = abs(unique(rnorm(n=1, mean=m, sd=s))) – sara connor

2 Answers

1
votes

Here are two ways - depending if you wanted each group to not contain points from any other group.

I'll just use a dummy data frame that just has columns x, y, and size.

library(tidyverse)

dt <- tibble(x = runif(100), y = runif(100), size = runif(100))

Allowing groups to share the same points

Create a vector for the size of each group.

sample_sizes <- 2^(seq_len(4) + 1)
sample_sizes
#> [1]  4  8 16 32

Randomly sample the data frame and add a group column.

sampled <- map_dfr(
  sample_sizes,
  ~sample_n(dt, .),
  .id = "group"
)

Plot using facets.

ggplot(sampled, aes(x, y, size = size)) +
  geom_point() +
  facet_wrap(~group)

plot1

Requiring groups to have different points

First, we need a way to generate four 1s, eight 2s etc. This can be done using log2 and some tricks.

groups <- floor(log2(seq_len(nrow(dt)) + 3)) - 1
groups
#>  [1] 1 1 1 1 2 2 2 2 2 2 2 2 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 3 4 4 4 4 4 4 4
#> [36] 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 4 5 5 5 5 5 5 5 5 5 5
#> [71] 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5 5

Shuffle this vector and add it as a column.

dt$group <- sample(groups)

Facet using this new column to generate the desired plots.

ggplot(dt, aes(x, y, size = size)) +
  geom_point() +
  facet_wrap(~group)

plot2

1
votes

First of all, the question's data creation code can be greatly simplified, rewritten with no loops at all. R is a vectorized language and the following will create a data frame with the same structure.

Don't forget to set the RNG seed, in order to make the results reproducible.

library(ggplot2)

set.seed(2020)    # make the results reproducible

sd <- rep(rep(mySD, each = 2), 10)
M <- rep(myMeans, 2*10)
x <- abs(rnorm(n = 40, mean = M, sd = sd))
y <- abs(rnorm(n = 40, mean = M, sd = sd))
size <- runif(40, 1, 25)
iter <- seq_along(x)
dt2 <- data.frame(x, y, size, M, sd, iter)
dt2$col <- c("blue", "red")

Now the plots. The following function accepts a data frame X as its first argument and a number of points to draw as the second one. Then plots n points chosen at random with color col and size (a continuous variable) size.

plot_fun <- function(X, n){
  Colors <- unique(X[["col"]])
  Colors <- setNames(Colors, Colors)
  i <- sample(nrow(X), n)
  g <- ggplot(X[i,], aes(x, y, size = size, color = col)) +
    geom_point() +
    scale_color_manual(values = Colors) +
    theme_bw()
  g
}

plot_fun(dt2, 8)

To plot several values for n, produce the plots with lapply then use grid.arrange from package gridExtra.

plot_list <- lapply(c(4,8,16,32), function(n) plot_fun(dt2, n))
gridExtra::grid.arrange(grobs = plot_list)

enter image description here

Individual plots are still possible with

plot_list[[1]]
plot_list[[2]]

and so on.

Another way is to use faceting. Write another function, plot_fun_facets assigning the number of points to a new variable in the sample data frames, n, and use that variable as a faceting variable.

plot_fun_facets <- function(X, n){
  Colors <- unique(X[["col"]])
  Colors <- setNames(Colors, Colors)
  X_list <- lapply(n, function(.n){
    i <- sample(nrow(X), .n)
    Y <- X[i,]
    Y$n <- .n
    Y
  })
  X <- do.call(rbind, X_list)
  g <- ggplot(X, aes(x, y, size = size, color = col)) +
    geom_point() +
    scale_color_manual(values = Colors) +
    facet_wrap(~ n) +
    theme_bw()
  g
}

plot_fun_facets(dt2, c(4,8,16,32))

enter image description here