Bug in stata runiform()?

[Matres]

Dear Statalist members,

To select all observations, per ID, in my panel data set, I've written the following code:

local idrandom round(runiform()*10) // generate random number to select unique ID
tab id if id == $idrandom, nolabel // show results of corresponding ID
display $idrandom // display the random number used

This results in the following output


The BvD ID number is an unique identifier per firm (10 years of observations per firm).

Therefore, it should return only one BvD ID number, in this case, 3

Strangely enough, the display $idrandom does give only one result.

Could someone explain to my why Stata 12 returns TRUE in the IF statement for more values that ID == 3?

Thanks!

Kind regards,

Matthijs Alderliefste

[Jeph Herrin]

First, this is the wrong forum, this is the Mata forum.

Second, I'll point out that $idrandom references a global macro, not the local macro assigned in the first line. The global must be assigned the name of a variable. Here's how I reproduced your problem:

clear
set obs 100
gen id=round(_n/10,1)
gen id2=round(_n/20,1)
global idrandom id2


local idrandom round(runiform()*10) // generate random number to select unique ID
tab id if id == $idrandom           // show results of corresponding ID
display $idrandom                   // display the random number used

hth,
Jeph

[Aljar Meesters]

As a third point, I would like to point out that round(runiform() * 10) does not what you probably want it to do. You probably want to select an integer from 1 to 10 included, all with equal probability. However, with round(runiform() * 10) you select a number from 0 up to 10 included, where 0 and 10 have a smaller probability of being chosen than 1 to 9 (only the uniform numbers from 0 to 0.05 - epsilon map to zero and 0.95 to 1 map to 10). You probably want to use ceil(runiform() * 10) or floor(runiform() * 10 + 1) if the fact that runiform() returns up to 1 - 2 ^ (-32) concerns you.

[Sergiy Radyakin]

Matthijs,

your global is a name of the variable. display will not show you more than one value of course, but rather than giving an error message it shows the first value of that variable.
instead you may want to use:

list $idrandom
display "$idrandom"

No error or problem or bug on Stata's side.

Best, Sergiy Radyakin