I'm using set.seed() but getting different answers in each run

Question

I thought if I used set.seed() inside a function then every time I ran that function the same seed would be used and I would get the same quasi random output. Take the following example:

my_fun <- function(n, v1, v2){
  set.seed = 42
  return(runif(n, v1, v2))
}
my_fun(1,2,3)
#> [1] 2.078126
my_fun(1,2,3)
#> [1] 2.918556
my_fun(1,2,3)
#> [1] 2.189768

I was expecting to get the same result every time I ran that function with the same inputs. Can you give me some education on why I don't?

What you are doing does not set the seed. It changes the set.seed function to be a number. — Mike Wise, May 08 '18 at 20:37
More precisely, it masks the `set.seed` function in the `base` namespace with an object in the local environment. Even with the code as-is, you can still do `set.seed(42)`, as R will "know" that you mean the function (and not the local variable); and you can always refer to it as `base::set.seed(42)` if you're paranoid (but it is unnecessary). — r2evans, May 08 '18 at 20:39
Isn't it [`random.seed(42)` in python](https://stackoverflow.com/a/11527011/3358272)? — r2evans, May 08 '18 at 20:40
@r2evans right after I typed that I realized that's no excuse at all... basically I'm just brain dead it seems — JD Long, May 08 '18 at 20:41
jumping between languages can bring out the ... brain-dead ... in all of us. — r2evans, May 08 '18 at 20:41
Running `rm(set.seed)` should should clear things up, just in case. — AkselA, May 08 '18 at 20:42
my billiards improves with 1.25 beers ... I wonder if my coding does, too — r2evans, May 08 '18 at 20:42
and I can't delete it because you all were so helpful... so my shame shall live in infamy. — JD Long, May 08 '18 at 20:45
Technically, your "shame" will live in infamy anyway for those of us with enough rep to see deleted questions. So ... you're public-enough no matter what you do :-) — r2evans, May 08 '18 at 20:47
Actually this is probably a good question to have stick around --- I have myself seen two other people make (and ask about) *exactly* this kind of error, so you will probably be doing some confused people a favor if you don't try to get rid of it. — Glen_b, May 09 '18 at 00:45

score 4 · Accepted Answer · answered May 08 '18 at 20:38

set.seed() is a function expecting a parameter equal to the value you want to seed the pseudorandom number generator(prng) with. The seed is the value used to start the number generation from. Most prng will use the current time as default, but when you pass it a seed you are determining the starting value and therefore all values to come after it as well.

So you need to call it like set.seed(42) to set your seed appropriately

Here is another question that gives a good response on what this function is actually doing https://stats.stackexchange.com/questions/86285/random-number-set-seedn-in-r

I'm using set.seed() but getting different answers in each run

1 Answers1