st: Re: Data manipulation: how to keep only consecutive obs in an unbalanced panel

[Roberto <terrymondo@onetel.net.uk>]

Dear Nick
thank you very much for your kind reply
I had actually looked at the suggested faq but I am not quite sure it is 
the answer to what I am after
See the thing is that I need Stata to "clean" the dataset from 
non-consecutive obs  for each firm.
What I can achieve using either
. tsset time
. gen run = .
. replace run = cond(L.run == ., 1, L.run + 1)
. egen maxrun = max(run)

or

tsspell, f(L.time == .)
by id: egen maxrun = max(_seq)

is to identify the maximum # of consecutive obs per firm.
Say for example that for a firm in my unbalanced panel I have observations 
for the years: 1991,1992,1997,1998,1999,2000,2001.
So this firm has N=7 and maxrun=5
What I would like to do is to find a command to drop 1991 and 1992 and to 
keep only the 5 consecutive observations.
Is there a way to do this?
Sorry if I was not clear before and thank you very much for your time
Regards
Roberto Mura
PhD candidate
University of York



At 15:01 11/10/2004, you wrote:

> I am not clear what the issue is here. The complaint
> seems to be that Stata does what you tell it to do,
> and not something else that you want.
>
> I'd recommend a look at
>
> How do I identify runs of consecutive observations in panel data?
> http://www.stata.com/support/faqs/data/panel.html
>
> by Vince Wiggins and friend, which may help.
>
> Nick
> n.j.cox@durham.ac.uk
>
> Carter Ivan Rees
>
> > There are duplicate commands already written for Stata, run help
> > duplicates.
> >
> > The following link has also come in handy for me in the past
> > if you are
> > looking to count the recurrence of a particular individual
> > across time.
> >
> > Your issue of consecutive years would require the code offered here to
> > be tweaked a bit.
> >
> > http://www.stata.co.uk/support/faqs/data/dups.html
>
> Roberto
>
> > I am currently trying to run a GMM estimation using the xtabond2
> > command.
> >
> > I have an unbalanced panel so I would want to manipulate my
> > dataset in a
> > way that Stata only keeps the continuous observations for
> > each firm For
> > example consider these 3 firms
> >
> > code    name    year    run    maxrun    _spell    _seq    _end
> > n
> >      N
> > 135540  A       1995     1          5            1           1  1
> > 1
> >      6
> > 135540  A       1997    1       5       2       1       0       2
> > 6
> > 135540 A        1998    2       5       2       2       0       3
> > 6
> > 135540 A        1999    3       5       2       3       0       4
> > 6
> > 135540 A        2000    4       5       2       4       0       5
> > 6
> > 135540 A        2001    5       5       2       5       1       6
> > 6
> > 900327 B        1991    1       7       1       1       0       1
> > 10
> > 900327 B        1992    2       7       1       2       0       2
> > 10
> > 900327 B        1993    3       7       1       3       1       3
> > 10
> > 900327 B        1995    1       7       2       1       0       4
> > 10
> > 900327 B        1996    2       7       2       2       0       5
> > 10
> > 900327 B        1997    3       7       2       3       0       6
> > 10
> > 900327 B        1998    4       7       2       4       0       7
> > 10
> > 900327 B        1999    5       7       2       5       0       8
> > 10
> > 900327 B        2000    6       7       2       6       0       9
> > 10
> > 900327 B        2001    7       7       2       7       1       10
> > 10
> > 134982
> > C        1997    1       1       1       1       1       1       3
> > 134982 C        1999    1       1       2       1       1       2
> > 3
> > 134982 C        2001    1       1       3       1       1       3
> > 3
> >
> >
> > with the tsspell command I have been able to identify the
> > maximum number
> > of consecutive observations per firms.
> > The problem is that if I write
> > keep if maxrun>=5
> > stata only drops firm "C" whereas I would want it to drop all the
> > non-consecutive observations in all panels so that the database would
> > look
> > like:
> >
> > code            name    year
> > 135540  A       1997
> > 135540  A       1998
> > 135540  A       1999
> > 135540  A       2000
> > 135540  A       2001
> > 900327  B       1995
> > 900327  B       1996
> > 900327  B       1997
> > 900327  B       1998
> > 900327  B       1999
> > 900327  B       2000
> > 900327  B       2001
> >
>
> *
> *   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/