Bookmark and Share

Notice: On April 23, 2014, Statalist moved from an email list to a forum, based at statalist.org.


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: st: Different results with the same code


From   Nick Cox <[email protected]>
To   "[email protected]" <[email protected]>
Subject   Re: st: Different results with the same code
Date   Mon, 10 Feb 2014 11:41:23 +0000

Thanks for the closure.

Anyone interested in the title should note that unexpected dependence
of results on sort order is one of the most common reasons for
different results on repeating code.
Nick
[email protected]


On 10 February 2014 10:41, Estrella Gomez <[email protected]> wrote:
> Sorry, I've realized that you were right: some variables in (firstnm)
> vary also by country and orig
>
> Thank you very much!
> Estrella
>
> 2014-02-10 11:24 GMT+01:00 Estrella Gomez <[email protected]>:
>>  Hi, Nick
>>
>> I do the collapse by country and orig; and the variables I include in
>> (firstnm) do not vary by country and orig, that's why I thougth that
>> picking the first value in each case would be enough; because it
>> should be equal to the rest of cases in that pair. Is not like that?
>>
>> Thanks,
>> Estrella
>>
>>
>> 2014-02-10 11:08 GMT+01:00 Nick Cox <[email protected]>:
>>> Wrong way round: -collapse- reports (firstnm) as it finds them; it's
>>> the sorting that includes some arbitrariness.
>>>
>>> But evidently you have different values on variables for which you
>>> select (firstnm), so why do you want to pick the first non-missing and
>>> what sort order do you want?
>>>
>>> Nick
>>> [email protected]
>>>
>>>
>>> On 10 February 2014 09:59, Estrella Gomez <[email protected]> wrote:
>>>> Hi, Nick
>>>>
>>>> I have tried sorting just before the collapse and it happens the same
>>>> thing. Does (firstnm) use random values each time then?
>>>>
>>>> Thank you,
>>>> Estrella
>>>>
>>>> 2014-02-10 10:41 GMT+01:00 Nick Cox <[email protected]>:
>>>>> Contrary to your claim, I would point at this:
>>>>>
>>>>> Which values are (firstnm) is likely to depend on the exact sort order
>>>>> of your data and you are sorting before you -collapse-.
>>>>>
>>>>> As Phil Schumm often explains, you could get consistent results by
>>>>> insisting on stable sorting, but that would be hiding a bug. There's
>>>>> evidently some arbitrariness that is better fixed directly.
>>>>> Nick
>>>>> [email protected]
>>>>>
>>>>>
>>>>> On 10 February 2014 09:07, Estrella Gomez <[email protected]> wrote:
>>>>>> Dear Statalisters
>>>>>>
>>>>>> I am running a code and I have realized that every time I run it the
>>>>>> results (the coefficients and the standard errors) are different. This
>>>>>> is the code:
>>>>>>
>>>>>> clear all
>>>>>> use "itunes_grav_song", clear
>>>>>> sort id country orig
>>>>>>
>>>>>> egen  nb_id=tag(id)
>>>>>> bysort orig: egen nr_orig=sum(nb_id)
>>>>>> bysort orig country: egen nr_dest=sum(available)
>>>>>> gen avail_share=(nr_dest/nr_orig)
>>>>>> collapse (count) nr_songs=id (firstnm) contig comlang ldist
>>>>>> avail_share, by(country orig)
>>>>>> encode country, gen(d_id)
>>>>>> encode orig, gen(o_id)
>>>>>> gen home=country==orig
>>>>>>
>>>>>> reg avail_share ldist comlang home i.d_id i.o_id
>>>>>>
>>>>>>
>>>>>> I have checked and after the collapse I obtain every time the same
>>>>>> dataset, so I don't know what can be the reason
>>>>>>
>>>>>> Thanks for any help!
>>>>>> Estrella
>>>>>> *
>>>>>> *   For searches and help try:
>>>>>> *   http://www.stata.com/help.cgi?search
>>>>>> *   http://www.stata.com/support/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/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/faqs/resources/statalist-faq/
*   http://www.ats.ucla.edu/stat/stata/


© Copyright 1996–2018 StataCorp LLC   |   Terms of use   |   Privacy   |   Contact us   |   Site index