I think the problem ist that interview dates vary, so you collect
information from a maximum of 23 months for each wave (Interview date
January until December).
You could reshape your data not only by person-year observations but by
person-month observations per row. Use an index starting with 1 in January
1999, use the same index for your interview dates and as far as I know
those unemployment spells are also indexed in a similar way in GSOEP.
Use the person-months rows to generate your aggregated variable
(unemployment months within the last 12 months before the interview date)
with time-series operators e.g., and then reshape it back to person-year
rows.
Johannes
----------------------
Johannes Geyer
Deutsches Institut f�r Wirtschaftsforschung (DIW Berlin)
German Institute for Economic Research
Department of Public Economics
DIW Berlin
Mohrenstra�e 58
10117 Berlin
Tel: +49-30-89789-258
Alekos Bibudis <[email protected]>
Gesendet von: [email protected]
03/02/2008 19:48
Bitte antworten an
[email protected]
An
[email protected]
Kopie
Thema
Re: st: Rearranging Variables
Hi again
Im sorry, my initial post was not really clear. I will try to explain
correctly what I am after.
My Data (Paneldata / wide format):
- individual Data on month of Interview from 2000 to 2006 (month00 -
month06) coded 1 to 12
- unemployment status for each month in the years 1999 to 2005 (Jan99 to
Dez05) coded 1/0
- ID of the interviewee.
Target:
12 variables (month1 to month12) holding unemployment information of the
12 months before the interview (for the interviewmonths Jan00 to Jan06).
for example: Interviewmonth = January 2000, the first variable (month1)
should contain the data from Dec 1999, the second the value of Nov 1999
... the last (month12) the value of Jan 1999
What I think would be the right way:
reshaping variables into the long format, this way I will have the
interview month in each wave as well as unemployment status each month
like this:
ID year month Jan Feb Mar
... Dez
1 1999 - 1 1 0 0
1 2000 3 1 1 0 0
1 2001 7 0 0 0 1
1 2003 9 1 1 1 0
...
code Idea for the first month before the inteview:
generate Monat1 = Dez(year-1) if month == 1;
replace Monat1 = Jan if month == 2;
replace Monat1 = Feb if month == 3;
replace Monat1 = Mar if month == 4;
replace Monat1 = Apr if month == 5;
replace Monat1 = Mai if month == 6;
replace Monat1 = Jun if month == 7;
replace Monat1 = Jul if month == 8;
replace Monat1 = Aug if month == 9;
replace Monat1 = Sep if month == 10;
replace Monat1 = Okt if month == 11;
replace Monat1 = Nov if month == 12;
There are two problems with that code I don't know how to solve:
1. Assumed Data is sorted by ID and year, how do I tell Stata to use the
value of a variable one column above the actual one (that is in the year
before the interview was conducted)?
In the example above, in 2000 January was the month of interview for
respondent 1, the month before the interview is Dec 1999 which is not
stored in the same colum as the interviewmonth-data (I tried to
represent that in the code-idea with "Dez(year-1)" which I know is not
correct).
2. In order to create variables for 12 months I'd have to type this code
like twelve times with an increasing suffixvalue of the new generated
variable (month1 to month12) and the month occupational status was
conducted varying. How can this be shortened?
I hope I made myself much clearer this time and that one of you might
know the trick :-)
Alekos
>Alekos,
>It would be helpful if you could explain why your commands do not
>work. One problem with your code is that "if month`x' = 1" should be
>"if month`x' == 1".
> Friedrich
> On Feb 1, 2008 3:49 PM, Alekos Bibudis
<[email protected]> wrote:
> > Hi everyone.
> >
> > Using data from the GSOEP I`d like to transform some variables
handling
> > with information on occupational status.
> > There is a variable for each wave and month of the year holding 0/1
> > coded information wether the person was unemployed at this point in
time
> > or not (84 variables from Jan99 to Dez05).
> > I additionally have information on the month and year the interview
was
> > conducted. That is for each wave a variable coded 1 (Jan) to 12
(Dec) (7
> > variables from month00 to month06).
> >
> > What I'd like to arrange is to generate twelve new variables
holding the
> > information of unemployment status of the twelve months BEFORE the
> > interview was conducted (Monat1 to Monat12) starting with
Interviewmonth
> > Jan00 and ending with Jan06.
> >
> > The logic would be:
> > generate a new yariable and give it the value of the month before the
> > interview
> > generate another new variable and give it the value of two months
before
> > the interview
> > ...
> > generate a last variable and give it the value of twelve months befor
> > the interview
> >
> > For 2006 there is no information on unemployment status, because it
will
> > be gathered retrospectivively in the next wave.
> >
> > Here's what I've tried, I know it doesn't work, but maybe it'll help
to
> > see what I'm after:
> >
> > #d;
> > version 9.1;
> > local y = 1999;
> > foreach x in 00 01 02 03 04 05 06{;
> > gen Monat1 = Dez`y' if month`x' = 1;
> > local y = `y' + 1;
> > };
> >
> > I'd appreciate every help I can get on this.
> >
> > --
> > ___________________________
> > Dipl.-Soz. Alekos Bibudis
> > "Betrieb und Besch�ftigung im Wandel"
> > SFB 580, Projekt B2
> > Friedrich-Schiller-Universit�t Jena
> >
> > Phone: +49-(0)371-6957616
> > Website: http://www.sfb580.uni-jena.de
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
*
* For searches and help try:
* http://www.stata.com/support/faqs/res/findit.html
* http://www.stata.com/support/statalist/faq
* http://www.ats.ucla.edu/stat/stata/
