limits of _n

[Ben Hoen]

Hi all,

I have a dataset with 22,534,283 records. I would like to create a simple unique id so I used the following: g id=_n.

When I examine the contents of this variable, I find that there are duplicates. See screenshot attached from results window and a paste is below.

the other odd issue is that id's format is float rather than long. Not a big deal (maybe) but it struck me as odd.

Any ideas how to resolve this issue?

Thanks, in advance,

Ben Hoen
Berkeley Lab

=================================================

. drop id

. g id=_n

. duplicates report id

Duplicates in terms of id

--------------------------------------
copies | observations surplus
----------+---------------------------
1 | 18216483 0
2 | 2 1
3 | 4317798 2878532
--------------------------------------

[eric_a_booth]

Perhaps:

g double id = _n

[Ben Hoen]

That works! as does

g long id = _n

.

Thanks!

[Nick Cox]

Eric makes the main point. The default default (repetition intended) variable or storage type for numeric values is float. You can set up your own default, but Stata doesn't look at expressions with numeric results to optimise the variable type.

Default storage types are explained alongside generate at help generate.

The problem may be that beyond beginner stage people assume that they understand generate. (Happens to most of us, I guess: I know that I didn't know about a relatively recent addition to describe -- which may mean added within the last decade -- until I re-read the help.) More optimistically, the default doesn't bite very often.

[Ben Hoen]

Thanks Nick.

Ben