Re: st: Carry forward

[David Kantor <kantor.d@att.net>]

At 12:21 AM 10/4/2013, Tina wrote:
> Hi All,
>
> I am having some problems with carryforward and I wanted to see if =
> anyone else knows of any problems with this command.
>
> I have a panel dataset that I have created and I want to use the =
> carryforward command to fill in missing values. =20
>
> . tsset id enroll
>
>       panel variable:  id (unbalanced)
>
>        time variable:  enroll, 1 to 3, but with gaps
>
>                delta:  1 unit
>
> . tsfill,full
>
> . replace nvisits_year =3D 0 if nvisits_year =3D=3D .
>
> (1517200 real changes made)
>
> . **fill in missing data**
>
> . bysort id: carryforward age, gen(age_1)
>
> age_1:  (1517200 real changes made)
>
>
> . bysort id: carryforward pay, gen (pay_cf)
>
> pay_cf:  (1516900 real changes made)
>
>
> . replace age_1 =3D age_1 + 1 if age=3D=3D.
>
> (1517200 real changes made)
>
>
>
> I have reloaded the raw data files and ran my exact same do file, =
> however I get different results every time I run the code:
>
>
> . tsset id enroll
>
>       panel variable:  id (unbalanced)
>
>        time variable:  enroll, 1 to 3, but with gaps
>
>                delta:  1 unit
>
> . tsfill,full
>
> . replace nvisits_year =3D 0 if nvisits_year =3D=3D .
>
> (1517200 real changes made)
>
> . **fill in missing data**
>
> . bysort id: carryforward age, gen(age_1)
>
> age_1:  (1517200 real changes made)
>
> . bysort id: carryforward pay, gen (pay_cf)
>
> pay_cf:  (1516901 real changes made)
>
> . replace age_1 =3D age_1 + 1 if age=3D=3D.
>
> (1517200 real changes made)
>
> =20
> And the third time I run the code:
>
> . bysort id: carryforward pay, gen (pay_cf)
> pay_cf:  (1516894 real changes made)
>
>
> Any help on what might be happening to the datafile and why the changes =
> are not the same each time I run the code?
>
> Thanks
>
> Tina

Does -id- uniquely identify the observations? (Does sorting on it 
yield a unique order?) I don't think so. Based on what you have 
written, you need to combine it with -enroll-.
Thus,
bysort id (enroll): carryforward pay...

Actually, you need to be sure that the combination of the two 
variables uniquely identifies observations. For that, you can use 
-assertky- (from SSC) or -isid- (standard Stata). Thus,
assertky id enroll
or
isid id enroll, sort

--followed by
by id (enroll): carryforward pay...

Note that in this situation, you can just use -by-; no need for -bysort-.

Generally, -carryforward- requires a "secondary" set of sorting 
variables to obtain consistent and meaningful results.
This is stated in the help. But your inquiry has prompted me to add a 
warning feature into -carryforward-.
HTH
--David

*
*   For searches and help try:
*   http://www.stata.com/help.cgi?search
*   http://www.stata.com/support/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/