使用set.seed函数生成可重现的结果

Question

If you want to generate 50 reproducible samples, should you use set.seed function in the first way or the second way?

set.seed(1000)
for(i in 1:50)
  sample(x,10)

or

for(i in 1:50)
  set.seed(1000)
  sample(x,10)

Answer 1

First way

the first sets a starting point for a string of pseudo random numbers to be used in the loop.

the second way sets the same starting point at each run of the loop and will give the same number giving the same result 1:50 times.

say set.seed(1000) gives a set of random numbers (1,2,3,4,5...,50)

then in the first example you will get a sample corresponding to 1,2,3....,50

but in the second example you will get a sample corresponding to 1,1,1....,1

Answer 2

The first method will work if you run all lines at the same time. If you ran the loop a second time without resetting the seed to 1000, or if you had another random event between setting the seed and sampling then you will get different answers.

The second method doesn't work as it will return an identical sample set 50 times.

For reproducible samples I would do it this way, (I have set the number of samples, x, equal to 10)

x=10
for(i in 1:50){
  set.seed(i)
  print(sample(x,10))
}

This will give you 50 different sample sets that will also be exactly the same next time you run the code