Hey,
after running both solution and rechecking my results I need to
restate that the award goes to Nick with -panelthin-!
-panelthin- does create events in a panel framework which are at
least 20 days apart and therefore captures "multiple observations".
Thanks to both of you!
Best,
Kaspar
2009/11/25 Martin Weiss <[email protected]>:
>
> <>
>
> " I have just run Martin's suggestion and it work fine- it provides
> exactly what I need."
>
>
>
> Thanks for the flowers! Just to be sure, I only addressed the -by- issue in my reply, not the overlap between more than two events.
>
>
>
> HTH
> Martin
>
>
> -----Ursprüngliche Nachricht-----
> Von: [email protected] [mailto:[email protected]] Im Auftrag von Kaspar Dardas
> Gesendet: Mittwoch, 25. November 2009 10:14
> An: [email protected]
> Betreff: Re: st: AW: Creating non-overlapping events
>
> Hi guys (Martin & Nick),
>
> thanks to both of you!
>
> I have just run Martin's suggestion and it work fine- it provides
> exactly what I need. I looked into -panelthin- and it should give me
> the same results. I did not run it though.
>
> Best,
>
> Kaspar
>
>
>
> 2009/11/25 Nick Cox <[email protected]>:
>> At first sight, it looks as if this is a problem solved by -panelthin- on SSC.
>>
>> The existence of two dates is not problematic, as they are a fixed time apart.
>>
>> As I understand it, Kaspar needs to declare his data to be panel data with identifier something like
>>
>> egen id = group(companyname name)
>> tsset id eventdate
>>
>> and then apply -panelthin-.
>>
>> Nick
>> [email protected]
>>
>> Kaspar Dardas
>>
>> thanks for your promt answer. However, this won't solve my problem.
>> Since I did not describe the problem properly its all on me. Your
>> solution is perfect for what I have described.
>>
>> First, I use the -by- groups companyname AND name. I need both since
>> the same name can appear in a different company,as well. I thought
>> that implementing by groups would be easier but with your solution it
>> not possible- or, I dont know how to handle it...
>>
>> Furthermore, a situation might appear where two overlaps exist. (see
>> trade 6 and 7). Trade 6 and 7 have to be dropped. However trade 8 is
>> fine since it does not interfere with trade 5. I dont think your
>> solution will caputre this, or will it?
>>
>>> 1 "Company A" "Margetts, Rob" 14-Feb-03 06-Mar-03
>>> 2 "Company A" "Margetts, Rob" 17-Feb-03 09-Mar-03
>>> 3 "Company A" "Margetts, Rob" 14-Nov-03 04-Dec-03
>>> 4 "Company A" "Margetts, Rob" 14-May-04 03-Jun-04
>>> 5 "Company B" "Michael, Ham" 06-Aug-04 26-Aug-04
>>> 6 "Company B" "Michael, Ham" 07-Aug-04 27-Aug-04
>>> 7 "Company B" "Michael, Ham" 15-Aug-04 02-Sep-04
>>> 8 "Company B" "Michael, Ham" 28-Aug-04 08-Sep-04
>>
>>
>> 2009/11/24 Martin Weiss <[email protected]>:
>>
>>> No sign so far of which -by- groups you want, although I can see that they are relevant here. For the simple example, this is possible code:
>>>
>>>
>>>
>>> *************
>>> clear*
>>>
>>> inp byte trade companyname:mylab name:mylab2 /*
>>> */ str10(eventdate postdays), auto
>>> 1 "Company A" "Margetts, Rob" 14-Feb-03 06-Mar-03
>>> 2 "Company A" "Margetts, Rob" 17-Feb-03 09-Mar-03
>>> 3 "Company A" "Margetts, Rob" 14-Nov-03 04-Dec-03
>>> 4 "Company A" "Margetts, Rob" 14-May-04 03-Jun-04
>>> 5 "Company A" "Margetts, Rob" 06-Aug-04 26-Aug-04
>>> end
>>>
>>> compress
>>>
>>> gen eventdate2=date(eventdate, "DM20Y")
>>> format eventdate2 %tdMonth_DD,_CCYY
>>> gen postdays2=date(postdays, "DM20Y")
>>> format postdays2 %tdMonth_DD,_CCYY
>>>
>>> drop eventdate postdays
>>>
>>> gen byte overlap= /*
>>> */ eventdate2 < postdays2[_n-1] /*
>>> */ in 2/`=_N'
>>>
>>> //not just "overlap" as missing in first obs
>>> drop if overlap==1
>>>
>>> list, noobs
>>> *************
>>
>> Kaspar Dardas
>>
>>> I have the following problem. I need to create a non-overlapping event
>>> data set. I have two date variables 1st: "eventdate" which is the
>>> start of the event and 2nd: "postdays", which is the end of the event.
>>> The length of the event is exactly 20 days (which is, of course, the
>>> difference between "eventdate" and "postdays"). If two events overlap
>>> (which happens in trade 1 and 2, in the example below) I only take
>>> the first event and drop the 2nd (trade 2).
>>> How can I do this in Stata? I have about 140000 trades, several
>>> thousands companies etc....
>>>
>>> trade companyname name eventdate
>>> postdays
>>> 1 Company A Margetts, Rob 14-Feb-03
>>> 06-Mar-03
>>> 2 Company A Margetts, Rob 17-Feb-03
>>> 09-Mar-03
>>> 3 Company A Margetts, Rob 14-Nov-03
>>> 04-Dec-03
>>> 4 Company A Margetts, Rob 14-May-04
>>> 03-Jun-04
>>> 5 Company A Margetts, Rob 06-Aug-04
>>> 26-Aug-04
>>>
>>
>> *
>> * 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/
>>
>
> *
> * 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/
>
>
> *
> * 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/
>
*
* 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/
