CODEWITHSUNDEEP
rmontecarloresamplingchi-squared
user2976477
user2976477

Reputation:

Simulating samples in R, storing them and calculating test statistic

I used a Monte carlo algorithm to generate data samples of size 100 of a geometric distribution using inversion sampling:

gi.cdf.geom <- function(p,u){
k <- c()
k <- ceiling(log(1-u)/log(1-p)) - 1
return(k)
}

The function above is the inverse of the CDF of a geometric distribution

u1 <- runif(100)
gen.gi.cdf1 <- gi.cdf.geom(50/239,u1)
as.data.frame(table(gen.gi.cdf1))

What I do not know how to do is randomly simulate a 1000 data samples of size 100 and to calculate the chi-square test statistic for each sample. My attempt at creating the samples is the following:

for(i in 1:1000){
 n=100
 p=50/239
 {
  u=runif(n)
  values <- gi.cdf.geom(p,u)
 }
 print(values)

}

However this gives me all the samples of my console with no way of referring to them later.

I would really appreciate some help.

Thank you

Upvotes: 1

Views: 584

Answers (1)

James
James

Reputation: 66834

Use replicate. For example:

(x <- replicate(3,rgeom(10,50/239)))
      [,1] [,2] [,3]
 [1,]    5    3   12
 [2,]   15    2    3
 [3,]    5    5    0
 [4,]    4    2    1
 [5,]   13    0    8
 [6,]    0    3    0
 [7,]    3    1    6
 [8,]    0    6    2
 [9,]    0    4    4
[10,]    8    4    1

You can test on them using apply

apply(x,2,chisq.test)
[[1]]

        Chi-squared test for given probabilities

data:  newX[, i] 
X-squared = 47.566, df = 9, p-value = 3.078e-07


[[2]]

        Chi-squared test for given probabilities

data:  newX[, i] 
X-squared = 10, df = 9, p-value = 0.3505


[[3]]

        Chi-squared test for given probabilities

data:  newX[, i] 
X-squared = 37.3243, df = 9, p-value = 2.303e-05


Warning messages:
1: In FUN(newX[, i], ...) : Chi-squared approximation may be incorrect
2: In FUN(newX[, i], ...) : Chi-squared approximation may be incorrect

Upvotes: 2

Related Questions