Handling leap years in Stata is also documented in
http://www.stata.com/support/faqs/data/leapyear.html
It isn't really a frequently answered question, more a fairly awkward
question.
Nick
[email protected]
Jeph Herrin
I'm not sure that what you doing is correct, but
here is how to get the random date you request:
gen daysinyear=doy(mdy(12,31,year))
gen rndday=int(daysinyear*runiform())
gen rnddate=mdy(1,1,year)+rndday
the first line is to capture the days of year in case
it is a leap year.
[email protected] wrote:
> I work with Stata 10 (upgrate for Stata 11 is ordered) and I would
> appreciate your advices for a "random date generator".
>
> I have panel Data were survivaltime is a dependent variable.
>
> Excerpt from my datastructure:
> ID Year survivaltime status "born" date
for
> status 1 date f. st.2 date f. st.3
deliverydate
> 1 2000 1098 0 03Jan2000
10jan2003
> . . 18jun2009
> 1 2001 1098 0 03Jan2000
10jan2003
> . . 18jun2009
> 1 2002 1098 0 03Jan2000
10jan2003
> . . 18jun2009
> 1 2003 1098 1 03Jan2000
10jan2003
> . . 18jun2009
>
> 2 2000 3453 0 04Jan2000 .
.
> . 18jun2009
> 2 2001 3453 0 04Jan2000 .
.
> . 18jun2009
> 2 2002 3453 0 04Jan2000 .
.
> . 18jun2009
>
> 3 2000 3453 0 04Jan2000 .
.
> . 18jun2009
> 3 2001 3453 0 04Jan2000 .
.
> . 18jun2009
> 3 2002 3453 3 04Jan2000 .
.
> . 18jun2009
>
>
> The variable status has the numbers 0,1,2,3 and they mean 0=pending,
and
> 1,2,3 = "death". If the person "dies" because of reason 1 I have
> information in the colum "date for status 1" and date f.st.2 and date
f.
> st.3 are missing. The same system is for status 2 and 3. If the person
> does not die, the status in the last observed year is 0 and i have no
> dates.
>
> The survivaltime I computed with:
> replace survivaltime=date for status 1-"born" if (date for status 1!=.
&
> status!=.) etc
> and for the pendings:
> replace survivaltime=deliverydate-"born" if (date for status 1==. &
date
> f. st.2==. & date f. st.3==.)
>
> As you can see I computed the survivaltime not depending on the status
> (because I did not know how to work around the change from 0 to a
status
> of 1-3) but I assumed that if I have Status 1 there has to be a date
in
> "date for status 1".
>
> And here is my problem, because, as you can see at ID 3, sometimes I
have
> a status (in this case 3) but I do not have date information in "date
f.
> st.3". Because of this my survivaltime is computed wrong.
>
> Because I know the year when the person "died" I would like to
randomly
> generate a "death date" between 01.01. and 31.12. in the death year
for
> the correct colum? (So that if the person dies because auf reason 3 I
get
> a date in colum "date f st.3" and not only for the year of death (in
case
> of ID3 2002) but also the same date in 200 and 2001, as seen in the
> excerpt.
>
> Could somebody help my how to do this?
>
> Also I would like to know how to change the pending state in the years
> before the death. So that I do not have
>
> ID Year survivaltime status but have ID
Year
> survivaltime status
> 1 2000 1098 0 1 2000
> 1098 1
> 1 2001 1098 0 1 2001
> 1098 1
> 1 2002 1098 0 1 2002
> 1098 1
> 1 2003 1098 1 1 2003
> 1098 1
*
* For searches and help try:
* http://www.stata.com/help.cgi?search
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
