Re: st: ranking variables on the basis of total values of observations

[Nick Cox <njcoxstata@gmail.com>]

This is your problem and you can make it as difficult as you wish.

What I recommend is that you calculate whatever you want as totals and
then make those totals data (meaning, values of variables) before
passing them to -rowranks-. For example, you can use -sum()- to get
cumulative sums and then pass those you want to -rowranks- or you can
-collapse- first.

Note that nothing stops you calling -rowranks- with just a single observation.

On the other hand it is open to you to write your own program and/or
modify -rowranks- so that everything is included in your own code.

Nick

On Thu, Mar 29, 2012 at 3:02 PM, tashi lama <ltashi32@hotmail.com> wrote:
>
> Yes, it does look like a rowranks problem but i think it is lil deeper because I have a time series data and I want to be able to provide time range. So, if you look at the following dataset, rowranks will give me a rank for each row.. But that is not exactly what I want. I want to be able to rank not only for 05jan2010, 06jan2010 and so on but 05jan2010-08jan2010. Now, in order to this, I have to be able to sum the values from 05jan2010-08jan2010 for all var1 var2 and var3 like given below. If I can do that, there would be only one row where obs are the sum. If i can do that, I can use rowranks. So,
>
>
>
> date                     totalvar2 totalvar3 totalvar4
>
> 05jan2010-08jan2010           10      10       22
>
>
>
>
>
> Yes, now I could use rowranks. Ahh...actually this analogy might make things clear. You know how when you graph pie, you could give do sth like
>
>
>
> graph pie var1 var2 if tin(05jan2010, 08jan2010), plabel(_all percent) => stata will sum the values of obs for both var1 and var2 for that period and compute the percentage. Rowranks doesn't allow ifs looking at the syntax.
>
> Let me know if I could make any clearer.
>
> Thanking and really appreciating your time and wisdom,

>> Date: Wed, 28 Mar 2012 23:22:15 +0100
>> Subject: Re: st: ranking variables on the basis of total values of observations
>> From: njcoxstata@gmail.com
>> To: statalist@hsphsun2.harvard.edu
>>
>> If I understand your problem correctly, it is that of -rowranks-.
>>
>> See also for a review in that territory
>>
>> SJ-9-1 pr0046 . . . . . . . . . . . . . . . . . . . Speaking Stata: Rowwise
>> (help rowsort, rowranks if installed) . . . . . . . . . . . N. J. Cox
>> Q1/09 SJ 9(1):137--157
>> shows how to exploit functions, egen functions, and Mata
>> for working rowwise; rowsort and rowranks are introduced
>>
>> This is now accessible to all at
>> http://www.stata-journal.com/sjpdf.html?articlenum=pr0046
>> under the Stata Journal's three-year window.
>>
>> Nick
>>
>> On Wed, Mar 28, 2012 at 9:44 PM, tashi lama <ltashi32@hotmail.com> wrote:
>> > Hello Nick,
>> >
>> > My apology, I mean rowranks, not rowsum. So, let's see if I could simplify it...
>> >
>> >
>> >
>> > Essentially I want to be able to rank var2 in the following dataset looking at its total values of obs from 05jan2010 to 09jan2010. so, I would use generate sum to get a rumsum. That would look like
>> >
>> >
>> >
>> >
>> >
>> > date var2 var3 var4 totalv~2 totalv~3 totalv~4
>> > 1. 05jan2010 3 1 7 3 1 7
>> > 2. 06jan2010 2 3 6 5 4 13
>> > 3. 07jan2010 4 4 5 9 8 18
>> > 4. 08jan2010 1 2 4 10 10 22
>> > 5. 09jan2010 5 8 3 15 18 25
>> > 6. 10jan2010 8 9 2 23 27 27
>> > 7. 11jan2010 4 6 3 27 33 30
>> > 8. 12jan2010 3 3 8 30 36 38
>> > 9. 13jan2010 1 2 3 31 38 41
>> >
>> >
>> >
>> >
>> > So, I would look at the observations at 09jan2010. So, as you can see, var2 has 5, var3 has 8 and so on. since 5 is the second lowest in that row after 3, I would say var2 is 2nd. Now, I have to able to code this and that is exactly I think you guys can help me or give me a lead.
>> >
>> >
>> >
>> > Thank you and let me know if I could make more clea.

>> >> Date: Wed, 28 Mar 2012 21:17:19 +0100
>> >> Subject: Re: st: ranking variables on the basis of total values of observations
>> >> From: njcoxstata@gmail.com
>> >> To: statalist@hsphsun2.harvard.edu
>> >>
>> >> I don't know what -rowsum- is here.
>> >>
>> >> Thanks, but this doesn't get me much closer. Of course, if other
>> >> people can work out what you want, they should chip in.
>> >>
>> >> Please _show_ what you want it changed to, as I asked earlier. No
>> >> word description, just what the dataset would like look after changes.
>> >>
>> >> Nick
>> >>
>> >> On Wed, Mar 28, 2012 at 8:49 PM, tashi lama <ltashi32@hotmail.com> wrote:
>> >>
>> >> > date var2 var3 var4
>> >> > 1. 05jan2010 3 1 7
>> >> > 2. 06jan2010 2 3 6
>> >> > 3. 07jan2010 4 4 5
>> >> > 4. 08jan2010 1 2 4
>> >> > 5. 09jan2010 5 8 3
>> >> > 6. 10jan2010 8 9 2
>> >> > 7. 11jan2010 4 6 3
>> >> > 8. 12jan2010 3 3 8
>> >> > 9. 13jan2010 1 2 3
>> >> >
>> >>
>> >> > Say, I would like to rank var2 for certain period of time say (05jan2010-09jan2010). So, I would have Stata add the obs values from 05jan2010 to 09jan2010 for all the variables and rank var2 like 1st or 2nd or 3rd. I looked at rowsum. I don't think it will help.
>> >>
>> >> >> Date: Wed, 28 Mar 2012 20:34:58 +0100
>> >> >> Subject: Re: st: ranking variables on the basis of total values of observations
>> >> >> From: njcoxstata@gmail.com
>> >> >> To: statalist@hsphsun2.harvard.edu
>> >> >>
>> >> >> As far as I can see this could mean several things. I am not clear
>> >> >> that when you say "rank" you don't mean "order" instead.
>> >> >>
>> >> >> I don't think you want -rowranks- (SSC).
>> >> >>
>> >> >> Why not just give us a toy dataset, with say 5 observations and 5
>> >> >> variables, and what you want to change it to? Then it should be easier
>> >> >> to see what you want.
>> >> >>
>> >> >> By the way, adding lots of blank lines just makes your posts more
>> >> >> difficult to read. I've edited what came in.
>> >> >>
>> >> >> Nick
>> >> >>
>> >> >> On Wed, Mar 28, 2012 at 8:22 PM, tashi lama <ltashi32@hotmail.com> wrote:
>> >> >> >
>> >> >> > I have a list of variables and I would like to rank the variables or any given variable for that matter on the basis of total values of their observations. I thought of approaching this problem in the following way.
>> >> >> >
>> >> >> > 1. find the running sum of all the variables using generate or the total using egen although I think I would prefer generate.
>> >> >> >
>> >> >> > say there are var1 var2 var3
>> >> >> >
>> >> >> > gen tvar1=sum(var1)
>> >> >> > gen tvar2=sum(var2)
>> >> >> > gen tvar3=sum(var3)
>> >> >> >
>> >> >> > 2. then compare tvar1, tvar2 and tvar3 using if conditions.
>> >> >> >
>> >> >> > Once this is done, I would like to expand such that I can rank those variables but for given period of time. And this is essentially why I think generate sum is better for this problem because it is a running sum.
>> >> >>
>> >> >> > say date var1 var2 var3
>> >> >> >
>> >> >> > I would like to be able to rank variables say for 01jan2011 to 01feb2011 and so on.
>> >> >> >
>> >> >> > Has anyone worked in this kind of problem before or does anyone have any idea or thought regarding this problem? Any help or lead would be highly appreciated? I saw that there is a stata module rowranks to calculate ranks in a row but i don't know how can it be useful to me. Most probably not....

*
*   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/