Search code examples
rrandomsamplingresampling

Sampling in the wrong direction R

Background

I have two variables called x and y (please see R code below the picture). When I plot(x, y), I obtain the top-row plot (see below). y values are stacked over the top of each x value. I then try to sample from these y values and make a second plot below the mother plot.

Question

I am wondering WHY when I use the predit.range (see R code below) to be 10:0 (the problem doesn't happen when I use 0:10) my sampling procedures goes completely in the wrong direction? (please compare the top-row plot to the bottom-row plot)

enter image description here

############# Input Values ################
                      each.sub.pop.n = 150; 
                      sub.pop.means = 20:10; 
                      predict.range = 10:0; 
                      sub.pop.sd = .75;
                      n.sample = 2;
#############################################
par( mar = c(2, 4.1, 2.1, 2.1) )

m = matrix( c(1, 2), nrow = 2, ncol = 1 ); layout(m)

Vec.rnorm <- Vectorize(function(n, mean, sd) rnorm(n, mean, sd), 'mean')

y <- c( Vec.rnorm(each.sub.pop.n, sub.pop.means, sub.pop.sd) )

x <- rep(predict.range, each = each.sub.pop.n)

plot(x, y)

## Unsuccessful Sampling ## The problem must be lying in here:

sampled <- lapply(split(y, x), function(z) sample(z, n.sample, replace = TRUE))
sampled <- data.frame(y = unlist(sampled), 
                 x = rep(predict.range, each = n.sample))
plot(sampled$x, sampled$y)
Solution

  • This is sufficient to illustrate why.

    x <- 10:0; y <- 10:0

    Did you notice how

    split(y, x)

    sorts the list? To get your desired ordering, control factor levels:

    split(y, factor(x, levels = unique(x))

    In your context, you can use efficiently without unique:

    split(y, factor(x, levels = predict.range))