How to write a while statement for the purpose detailed in the post?

[Xueting Wang]

Dear all,

I have two variables: seq1 and u.

seq1 is a repeating sequence: 1,2,3,4,5
u is a random integer between 1 and 5, generated by u = 4*uniform() + 1

Now I want to achieve this: when u equals to seq1, do the randomisation again until u has a different value as seq1.

My code is:

foreach var in u{
foreach num in seq1{
while `num' == `var'{
replace `var' = 4*uniform() + 1)
}
}
}

No error in this code but it didn't work for my purpose. After I run it, nothing changed.

So would you mind help me figure out why this code cannot work and what I need to do to achieve my purpose?

Thank you very much!

[Joro Kolev]

I think simply doing

replace  u = 4*uniform() + 1 if u==seq1

should achieve the goal you have in mind.

[Xueting Wang]

Thanks for the suggestion.

Yes, it indeed works but I have to manually run it several times in order to eliminate all repeating values in 2 columns. I'm wondering whether using while loop could make this process automatic?

Look forward to alternative solutions.

[William Lisowski]

I'm not sure I agree with Joro, but that's largely because I'm totally unsure what you are trying to achieve.

u is a random integer between 1 and 5, generated by u = 4*uniform() + 1

No, u is not a random integer between 1 and 5. The uniform() function generates a random real number between 0 and 1. To generate a random integer between 1 and 5, you use the function runiformint(1,5).

I do not understand what

when u equals to seq1, do the randomisation again until u has a different value as seq1.

Is seq1 also a variable in your dataset, taking the values 1 2 3 4 5 1 2 3 4 5 and so forth? So if in the third observation seq is 3 and u is also 3, you want to change u to one of 1 2 4 5, and if u is not 3, you want to leave it alone?

In that case

generate temp = runiformint(1,4)
replace u = temp + cond(temp<seq1,0,1) if u==seq1
drop temp

will do what you want. Or, you could use the following code to generate u to begin with:

generate u = runiformint(1,4)
replace u = u + cond(u<seq1,0,1)

[Joro Kolev]

Indeed there is some misunderstanding here. If you are doing what the command you show is doing, that is generating a continuous uniform variable u~(1,5), theoretically speaking the probability of this u coinciding with a set of 3 discrete values 2,3,4 is 0.

In 1 million observations, I got 0 coincidences. So how many times you might possibly need to re-randomise if by some chance you got a coincidence?

. set obs 1000000
number of observations (_N) was 0, now 1,000,000

. set seed 69

. egen seq1 = seq(), from(1) to(5)

. gen u = 4*uniform() + 1

. assert seq1!=u

. summ u

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
           u |  1,000,000    2.999572    1.155054    1.00001   4.999993

. summ seq1

    Variable |        Obs        Mean    Std. Dev.       Min        Max
-------------+---------------------------------------------------------
        seq1 |  1,000,000           3    1.414214          1          5

.