Correctly Writing a Double Loop in R

General
1

I have the following dataset:

library(dplyr)
library(purrr)
set.seed(123)
my_data1 = data.frame(var1 =  rnorm(500,100,100), prop = runif(500, min=0, max=0.5))
my_data2 = data.frame(var1 = rnorm(500, 200, 50), prop = runif(500, min=0.4, max=0.7))
my_data = rbind(my_data1, my_data2)

I made ten equal-sized bins, found the average proportion within each bin, then plotted the results:

final = my_data %>%
arrange(var1) %>%
mutate(ntile = ntile(var1, 10)) %>%
group_by(ntile) %>%
summarise(mean = mean(prop))

plot(final$ntile, final$mean, type = "l", xlab = "Bin Number", ylab = "Average Proportion", main = "Relationship Between Bins and Average Proportion")

image
image1361×562 7.09 KB

Now, suppose I:

  • Take a random 70% sample, create 10 bins and calculate the mean proportion within each of these bins
  • Then, from the remaining 30% sample - I again take a random sample and create 10 bins and calculate the mean proportion within each of the bins
  • Next, I take the squared sum of the difference between these two steps for each bin
  • Finally, repeat this process many times (while tracking the results)

Below, I wrote some R code for this procedure:

base = sample_n(my_data, 700)

base_comp = base %>%
            arrange(var1) %>%
            mutate(ntile = ntile(var1, 10)) %>%
            group_by(ntile) %>%
            summarise(mean = mean(prop))


sampling_frame = my_data %>% anti_join(base)

my_list = list()

for (i in 1:1000)
    
{
    
    a_i = sample_n(sampling_frame, 100)
    
    base_a_i = a_i %>%
    arrange(var1) %>%
    mutate(ntile = ntile(var1, 10)) %>%
    group_by(ntile) %>%
    summarise(mean = mean(prop))
    
    sum_i <- sum(map2_dbl(base_a_i$mean, base_comp$mean, function(x, y) (x - y)^2))
    
    my_list[[i]] <- data.frame(id = i, sum_i)
    print(data.frame(id = i, sum_i))
}

results = do.call(rbind.data.frame, my_list)

plot(density(results$sum_i), main = "Distribution of Deviations")
plot(results$id, results$sum_i, xlab = "Iteration", ylab = "Deviation", main = "Trace Plot", type = "b")

image
image1361×562 21.4 KB

My Question: Now, I am trying to create a "double loop" - that is, I would to:

  • For j = 1, randomly create "base_j" , "base_comp_j" and "sampling_frame_j"
  • Repeat the "i" loop 100 times
  • Take the average of all "sum_i" for j=1 (i.e. average_sum_i_j)
  • Now, repeat this for j = 2
  • Repeat until j = 100
  • Each point on these two graphs will be "average_sum_i_j"

Here is my attempt to do this:

#  change i and j index to 10 for brevity

my_list = list()

for (j in 1:10) {
    base_j = sample_n(my_data, 700)
    base_comp_j = base_j %>%
        arrange(var1) %>%
        mutate(ntile = ntile(var1, 10)) %>%
        group_by(ntile) %>%
        summarise(mean = mean(prop))
    sampling_frame_j = my_data %>% anti_join(base_j)
    sum_i_list = list()
    for (i in 1:10) {
        a_i = sample_n(sampling_frame_j, 100)
        base_a_i = a_i %>%
            arrange(var1) %>%
            mutate(ntile = ntile(var1, 10)) %>%
            group_by(ntile) %>%
            summarise(mean = mean(prop))
        sum_i <- sum(map2_dbl(base_a_i$mean, base_comp_j$mean, function(x, y) (x - y)^2))
        sum_i_list[[i]] <- sum_i
    }
    average_sum_i_j = mean(unlist(sum_i_list))
    my_list[[j]] <- data.frame(id_j = j, id_i = i, average_sum_i_j)
    print(data.frame(id_j = j, id_i = i, average_sum_i_j))
}

results = do.call(rbind.data.frame, my_list)

plot(density(results$average_sum_i_j), main = "Distribution of Deviations")
plot(results$id, results$average_sum_i_j, xlab = "Iteration", ylab = "Deviation", main = "Trace Plot", type = "b")

However, I don't think I am doing this right since the loop does not appear to be cycling through both indices.

Can someone show me how to correct this?

Thanks!

2

You can use nested loops to achieve this.

First, you can create a outer loop that runs from j = 1 to j = 100. Inside this outer loop, you can create a new random sample for "base_j", "base_comp_j" and "sampling_frame_j".

Then, you can create an inner loop that runs from i = 1 to i = 100, where you repeat the same process as before to calculate "sum_i" for each iteration.

Finally, you can take the average of all "sum_i" for a given j and store it in a variable "average_sum_i_j".

You can then plot the distribution of "average_sum_i_j" and "j" on a graph to visualize the results.

Here is an example of the modified code to achieve this:

Initialize an empty list to store the average sum_i for each j

average_sum_i_j_list <- list()

Outer loop to run from j = 1 to j = 100

for (j in 1:100) {

Copy code

# Create random samples for base_j, base_comp_j and sampling_frame_j
base_j <- sample_n(my_data, 700)
base_comp_j <- base_j %>%
    arrange(var1) %>%
    mutate(ntile = ntile(var1, 10)) %>%
    group_by(ntile) %>%
    summarise(mean = mean(prop))
sampling_frame_j <- my_data %>% anti_join(base_j)

# Initialize an empty list to store the sum_i for each i for a given j
sum_i_list <- list()

# Inner loop to run from i = 1 to i = 100
for (i in 1:100) {
    
    # Create a random sample of 100 from sampling_frame_j
    a_i <- sample_n(sampling_frame_j, 100)
    
    # Calculate base_a_i
    base_a_i <- a_i
1 Like
3

This topic was automatically closed 21 days after the last reply. New replies are no longer allowed.

If you have a query related to it or one of the replies, start a new topic and refer back with a link.