[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: st: add up variable / quantile
Nick Cox <email@example.com>
Re: st: add up variable / quantile
Thu, 14 Apr 2011 07:17:29 +0100
Good. I had in my mind "Why not use -xtile- anyway?", but sorry, that
didn't make it to my previous post.
On Wed, Apr 13, 2011 at 10:26 PM, Scharnigg, Stan (Stud. SBE)
> Thank you, that was indeed the problem. I solved it with the xtile command.
> Van: firstname.lastname@example.org [email@example.com] namens Nick Cox [firstname.lastname@example.org]
> Verzonden: woensdag 13 april 2011 22:07
> Aan: 'email@example.com'
> Onderwerp: RE: st: add up variable / quantile
> I don't follow this well, but I see that you are including a comparison with r(p75), which is presumably thought to be left over from some previous command.
> However, r-class results are ephemeral and don't stick around forever. In particular, they get overwritten by -egen-, which does its own internal -count- at some point.
> If, however, r(p75) is undefined, then it's treated as missing and your comparison would be whether values were missing or greater than that, which is evidently not true for your data.
> I rarely understand why people who have quantitative data want to degrade it to indicator variables. That sounds most unstatistical to me.
> That aside, I think you need to reissue whatever command produced the r(p75) before you try to use it.
> -----Original Message-----
> From: firstname.lastname@example.org [mailto:email@example.com] On Behalf Of Scharnigg, Stan (Stud. SBE)
> Sent: 13 April 2011 20:42
> To: firstname.lastname@example.org
> Subject: RE: st: add up variable / quantile
> I still have a problem with this.
> My goal is to identify the top quantile for each years (total 6 years) in different variables
> What I did so far:
> gen year=year(newdate) // create years instead of normal dates
> egen gross_performance_years=total(gross_performance), by(accountID year) // create gross_performance per year
> egen tag_year=tag(RekeningID year) // tag a 1 for each year per accountID
> this all works fine, however now I want to create the top quantile variable for each year. So i did the following:
> gen topq_2000=gross_performance_years >=r(p75) & year==y(2000), however this doesn't work. I only get "0" as value.
> I also tried this:
> generate topq_2000 = 0
> replace topq_2000 = 1 if gross_performance_year >=r(p75) & tag_year==1 & year==2002
> but without succes
> Does anybody has some tips how I can do this?
> Van: email@example.com [firstname.lastname@example.org] namens Nick Cox [email@example.com]
> Verzonden: dinsdag 29 maart 2011 12:44
> Aan: firstname.lastname@example.org
> Onderwerp: Re: st: add up variable / quantile
> On Tue, Mar 29, 2011 at 11:30 AM, Scharnigg, Stan (Stud. SBE)
> <email@example.com> wrote:
>> Look at the help for -egen-. You want
>> egen total_gross = total(gross), by(accountID)
>> Thank you, but I have some additional questions:
>> A. I have data for 6 years (72 months). What if I want to add up the gross_performance for e.g the first 12 months. So, I guess I need to
>> create different variables for different time periods, but I am not sure how to do that. One extensive possibility might be that I create a different dataset
>> for every period, but I guess there might also be another solution
>> accountID; gross_performance; date
>> 1 -.1 jan_00
>> 1 .2 febr_00
>> 3 .1 jan_00
>> 3 .1 febr_00
> You can specify -if- on an -egen- command. Different summaries will in
> general require new variables (but not new datasets).
>> B. If I use egen total_gross = total(gross_performance), by(accountID) I get many duplicate values. In some cases I have
>> 72 duplicate values. What is the best way to delete the duplicate values, so that they won't show up if I do some tests. I don't
>> think that renaming them to "0" is an option then.
> You can use
> egen tag = tag(accountID)
> and then add
> ... if tag
> to commands to ensure that each summary is used once only. You cannot
> delete (in Stata -drop-) without losing other information.
> Alternatively, -collapse- will yield a reduced dataset with one
> observation for each account.
* For searches and help try: