set.seed with R 2.15.2

Question

My understanding is that using set.seed ensures reproducibility but this is not the case with the following R code in R 2.15.2. Am I missing something here?

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875
 rnorm(5)
[1] -1.8179560  0.6300986 -0.2761841 -0.2841597 -0.9193220

Answer 1

set.seed() reinitializes the random number generator.

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875

set.seed(12345)
rnorm(5)
[1]  0.5855288  0.7094660 -0.1093033 -0.4534972  0.6058875

Answer 2

Any call that uses the random number generator will change the current seed, even if you've manually set it with set.seed.

set.seed(1)
x <- .Random.seed # get the current seed
runif(10) # uses random number generator, so changes current seed
y <- .Random.seed
identical(x, y) # FALSE

As @StephanKolassa demonstrates, you'd have to reset the seed before each use of the random number generator to guarantee that it uses the same one each time.