Re: st: RE: creating variable summarizing for each individual properties of other members of a group at t-1

[Nick Cox <njcoxstata@gmail.com>]

Thanks. I (think I) am understanding more. Now sounds like

0. Tag individuals' first entries

bysort ind_id (yearmonth) : gen ind_score = sum(ind_entry)
by ind_id : replace ind_score = ind_score == 1  & ind_score[_n-1] != 1

/// != 1 not == 0 to catch ind_entry == 1 when _n == 1 for which
ind_entry[_n-1] is missing

See also

FAQ     . . . . . . . . . . . . . . . First and last occurrences in panel data
        . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  N. J. Cox
        3/07    How can I identify first and last occurrences
                systematically in panel data?
                http://www.stata.com/support/faqs/data/firstoccur.html



1. Get the sum of all individual first entries

sort yearmonth
gen all_score = sum(score)
by yearmonth : replace all_score = all_score[_N]

2. Subtract this individual

replace all_score = all_score - ind_score

3. Look one step back in time

bysort ind_id (yearmonth) : gen prevscore = allscore[_n-1]

Nick

On Tue, May 17, 2011 at 7:38 PM, Erik Aadland <erikaadland@hotmail.com> wrote:
> My apologies if I am being imprecise.
>
> I am struggling in my attempt to create the following two variables.
>
> One variable that yields a score for each individual in the dataset for each yearmonth: the score is the sum of all ind_entry = 1 for all the other individuals excluding the focal individual up until and including time: yearmonth - 1.
>
> And a second variable that identifies for each individual in the dataset for each yearmonth: the total number of other individuals in the dataset that have experienced ind_entry = 1 at least once up until and including time: yearmonth - 1. In this variable, I want to count the number of individuals that have experienced ind_entry = 1, not how many times they have entered.

Nick

>> I naturally have no idea what -ind_entry- means.
>>
>> I am just summing it, as the code implies. Is that not what you want?

Erik Aadland

>> Thank you so much for your input, Nick.
>>
>> I have experimented and generated different variables previously relying on the very helpful FAQ in question.
>>
>> I am struggling with this problem however. When I apply the suggested code below, it appears that the calculation of the number of peers adds up more than it should for ind_ids with more ind_entry = 1 relative to other ind_ids, and consequently contribute more to the "score" than those with fewer ind_entry = 1.
>>
>> Referring to the example dataset, ind_id 2 is given the correct "prevscore". Ind_id 4, however, is not. By yearmonth 12, ind_id 4 has contributed 2 ind_entry=1 to the "score", which is correct for ind_id 2. However, ind_id 2 has not yet experienced ind_entry=1. Consequently, score - 1 for ind_id 4 yields a score = 1 in yearmonth 11 and 12, when the correct score = 0. And so on.
>>
>> Here is the suggested code as I applied it:
>>
>> clear ;
>> #delimit ;
>> use "ind_entry_ex.dta" ;
>> sort yearmonth ;
>> gen score = sum(ind_entry) ;
>> by yearmonth: replace score = score[_N] ;
>> replace score = score - ind_entry ;
>> bysort ind_id (yearmonth): gen prevscore = score[_n-1] ;
>>
>>
>> Here is the output:
>>
>> year month yearmonth ind_id ind_entry score prevscore
>> 2003 10 10 2 0 1
>> 2003 11 11 2 0 1 1
>> 2003 12 12 2 0 2 1
>> 2004 1 13 2 0 2 2
>> 2004 2 14 2 1 2 2
>> 2004 3 15 2 0 3 2
>> 2003 10 10 4 1 0
>> 2003 11 11 4 0 1 0
>> 2003 12 12 4 1 1 1
>> 2004 1 13 4 0 2 1
>> 2004 2 14 4 0 3 2
>> 2004 3 15 4 0 3 3
>>
>> I use Stata 10.

Nick

>> > I don't know what "I am familiar with" means here. Does it mean that you've read the FAQ but can't see how to apply it?
>> >
>> > This sounds to me like
>> >
>> > 1. Get the sum of all individual entries
>> >
>> > sort yearmonth
>> > gen score = sum(ind_entry)
>> > by yearmonth : replace score = score[_N]
>> >
>> > 2. Subtract this individual
>> >
>> > replace score = score - ind_entry
>> >
>> > 3. Look one step back in time
>> >
>> > bysort ind_id (yearmonth) : gen prevscore = score[_n-1]


Erik Aadland

>> > I need to create a variable that sums for each individual in my dataset the total number of ind_entry of all other individuals at time: yearmonth - 1.
>> > I have attached a small ex of my data structure below. So for instance, given the small dataset below, for ind_id 2 in yearmonth 11 this variable score = 1. But for ind_id 4 in the same yearmonth, the score = 0.
>> >
>> > I would also like to generate a variable that identifies for each individual the unique number of other individuals in the dataset that have experienced ind_entry = 1 at least once up until time: yearmonth - 1.
>> >
>> > I am familiar with the following FAQ: http://www.stata.com/support/faqs/data/members.html
>> >
>> > My data structure is snapshot data in principle like the example below, but some individuals enter the observation window later than others (i.e. in later yearmonths):
>> >
>> > year month yearmonth ind_id ind_entry
>> > 2003 10 10 2 0
>> > 2003 11 11 2 0
>> > 2003 12 12 2 0
>> > 2004 1 13 2 0
>> > 2004 2 14 2 1
>> > 2004 3 15 2 0
>> > 2003 10 10 4 1
>> > 2003 11 11 4 0
>> > 2003 12 12 4 1
>> > 2004 1 13 4 0
>> > 2004 2 14 4 0
>> > 2004 3 15 4 0

*
*   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/