Initialize vector using the ifelse function in R

Question

I'm trying to run the next algorithm in R:

simulation <- function(n){
  f <- function(x){.7*h(x)+.3*g(x)}
  x <- vector("numeric", 2)
  w <- vector("numeric", 2)
  x0 <- c(3,3)
  for(i in 0:n){
    ifelse(i==0,w<-x0,w<-x) //After this line w=(3,3) (on the first iteration) 
    y <- rnorm(2, mean = w, sd = S1) //Also this line gives a non-zero value to y 
    alfa <- (f(y))/(f(w))
    ifelse(runif(1)<alfa,x<-y,x<-w) //Here the variables y and w are a two dimension zero vector. 
  }
  return(alfa[n])
}

I don't understand why at the second "ifelse" the vectors w and y are equal to (0,0). Maybe there is another way to initialize the vector x. But I don't get how the first "ifelse" works fine. Any suggestions would be great!

Could you add some expected output? Looks like the loop isnt necessary here — Donald Seinen, Nov 22 '21 at 14:00

score 0 · Answer 1 · answered Nov 22 '21 at 10:08

If I understood the problem correctly you are trying to simulate a vector containing 2 values which in the function is named as alfa.

Also in each iteration w and y are not equal (0, 0) it is updating as you have stated in the function. The above function only returns the nth value of alfa.

alfa only has 2 elements in each iteration and return(alfa[n]) with return value from the last iteration only.

I have created two simple function named h and g assuming you will also have similar ones and defined sd as 1.

Made a small tweak to your query to get n number of vectors containing 2 elements each.

h <- function(x) {
  x ^ 2
}

g <- function(x) {
  x ^ 3
}

simulation <- function(n, S1 = 1, seed = 123){
  f <- function(x){.7*h(x)+.3*g(x)}
  x <- vector("numeric", 2)
  w <- vector("numeric", 2)
  x0 <- c(3, 3)
  for(i in 0:n){
    ifelse(i==0, w <- x0, w <- x) # After this line w=(3,3) (on the first iteration)
    set.seed(seed)
    y <- rnorm(2, mean = w, sd = S1) # Also this line gives a non-zero value to y 
    alfa <- (f(y))/(f(w))
    set.seed(seed + 1)
    ifelse(runif(1) < alfa, x <- y, x <- w) # Here the variables y and w are a two dimension zero vector.
    print(alfa)
  }
}

simulation(3)

Output from above query

[1] 0.5917627 0.8156456
[1] 0.5236186 0.8027823
[1] 0.4268981 0.7878858
[1] 0.2798218 0.7704123

Or else if you want to store the simulation in alfa and then return the complete alfa without printing each iteration then you can use the below function which creates a matrix for alfa:

simulation <- function(n, S1 = 1, seed = 123){
  f <- function(x){.7*h(x)+.3*g(x)}
  x <- vector("numeric", 2)
  w <- vector("numeric", 2)
  x0 <- c(3, 3)
  alfa <- matrix(0, nrow = n + 1, ncol = 2) # matrix of zeros
  for(i in 0:n){
    ifelse(i==0, w <- x0, w <- x) # After this line w=(3,3) (on the first iteration) 
    set.seed(seed)
    y <- rnorm(2, mean = w, sd = S1) # Also this line gives a non-zero value to y 
    alfa[i + 1, ] <- (f(y))/(f(w))
    set.seed(seed + 1)
    ifelse(runif(1) < alfa[i + 1, ], x <- y, x <- w) # Here the variables y and w are a two dimension zero vector.
  }
  return(alfa)
}

simulation(3)

Output from the above query

          [,1]      [,2]
[1,] 0.5917627 0.8156456
[2,] 0.5236186 0.8027823
[3,] 0.4268981 0.7878858
[4,] 0.2798218 0.7704123

I have used alfa[i + 1, ] since indexing in R starts from 1 and not 0.

Initialize vector using the ifelse function in R

1 Answers1