Terminate while loop after conditions met ( R )

Question

I want to generate 50 numbers with rnorm, with two criteria. If there are numbers <99 or >101 (i.e., outside 99-101), omit them and run rnorm again until 50 total numbers that meet criteria are generated.

This is unrelated to statistics. I'm simply trying to learn how to use while loops.
I suspect the rnorm(50, part on line 3., is computationally inefficient -- any advice there would be great.
The main problem I have is that although this code works, it goes on for ever. It needs to terminate when there are 50 observations that meet criteria. Thus far, I have tried unsuccessfully to use if and break to do this...

Code so far:

1.   z = rnorm(50, mean = 100, sd = 10)
2.   while( match(TRUE, z > 99) > 0 | match(TRUE, z < 101) > 0 )  {
3.   z = c( z[z >= 99 & z <= 101], rnorm(50, mean = 100, sd = 10 )) 
4.   }

you can consider a random number generations with lower and upper bounds. See here http://stackoverflow.com/questions/19343133/setting-upper-and-lower-limits-in-rnorm — SabDeM, May 30 '15 at 04:02

Special Sauce · Accepted Answer · 2015-05-30T04:44:34.120

1

The reason your loop "never quits" in its current form is because the stats are literally stacked against you.

The probability of consecutively generating 50 normally distributed numbers within ±1 standard deviation of the mean is approximately 4.22E-9. And your tolerance is only 1/10^th of one standard deviation, so imagine how astronomically small your odds are.

The simplest way to enforce a fixed number of iterations of a loop is the for loop:

for (i in 1:50) {
  sum = sum + sum^0.5
}

Otherwise you can add a watchdog counter like follows:

z = rnorm(50, mean = 100, sd = 10)
wd = 0
while( match(TRUE, z > 99) > 0 | match(TRUE, z < 101) > 0 )  {
  z = c( z[z >= 99 & z <= 101], rnorm(50, mean = 100, sd = 10 ))
  wd = wd + 1
  if (wd == 50) { break }
}

Also keep in mind that loops in R are relatively slow (compared to more typical operations) and their use is discouraged unless there is a good reason. R is a functional language and most operators and functions are vectorized. You will find the vectorized operations perform substantially faster than equivalent loop-like / procedural flows.

edited May 30 '15 at 04:44

answered May 30 '15 at 03:57

Special Sauce

5,338
2
27
31

Thanks I was not trying to create 50 normally distributed numbers. Rather, I just wanted to see excluding occur for numbers outside 99-101 until I had 50 numbers. I have actually learned about watchdog counters before (e.g., in python)...why I didn't think of it here is a mystery to me. Anyway, thanks again for the help! – lnNoam May 30 '15 at 04:33
Ah, I see what you were trying to accomplish. Your current code won't work because the `c( )` command is doing a vector concatenation that is always adding 50 new normal numbers to `z`, so `z` is always growing in length. – Special Sauce May 30 '15 at 04:41
Yes, I fixed that. Thanks again for the help. – lnNoam May 30 '15 at 04:59

lnNoam · Answer 2 · 2015-05-30T05:29:35.677

I figured out this solution too, which will also do it:

rnorm_rg = function(x, mean, sd, lwr, upr){
  z = rnorm(x, mean = mean, sd = sd)
  while ( length(z[z >= lwr & z <= upr]) < x ) {
    z = c( z[z >= lwr & z <= upr], rnorm(x, mean = mean, sd = sd) )
    z = z[z >= lwr & z <= upr]
  }
  print(z)
}

e.g.,

rnorm_rg(  x = 50
           , mean = 100
           , sd = 10
           , lwr = 99
           , upr = 101
)

Terminate while loop after conditions met ( R )

2 Answers2