CODEWITHSUNDEEP
rfor-loopfunctional-programming
forRstudy
forRstudy

Reputation: 13

Summarising results of repeated hypothesis tests

I am generating two sets of data from normal distributions using rnorm(30, 10, 5). I am testing the hypothesis that the means are equal, using t.test, and repeating this process 1000 times to estimate the type 1 error rate.

My code is as follows:

for(i in 1:1000) {
  print(t.test(rnorm(30, 10, 5), rnorm(30, 10, 5),
               alternative="two.sided", var.equal=TRUE)$p.value < 0.05)
}

However, this returns countless TRUE and FALSE.

I can manually count the TRUEs, but is there an easier way?

Upvotes: 1

Views: 88

Answers (1)

jbaums
jbaums

Reputation: 27398

They're not countless. There are 1000 of them ;)

Instead of printing the result to screen, you should keep track of it in a vector.

A simple change will get you what you want:

sig <- logical(length=1000)
for(i in 1:1000) {
  sig[i] <- t.test(rnorm(30, 10, 5), rnorm(30, 10, 5),
               alternative="two.sided", var.equal=TRUE)$p.value < 0.05
}

Now you can tabulate the results:

table(sig)
## sig
## FALSE  TRUE 
##   956    44

A simpler approach would be to use replicate:

table(replicate(1000, t.test(rnorm(30, 10, 5), rnorm(30, 10, 5), 
  alternative="two.sided", var.equal=TRUE)$p.value < 0.05))

Upvotes: 2

Related Questions