Finding Values in a Column that are "Close"

[Jack Hays]

I have panel data that looks like the following:

* Example generated by -dataex-. For more info, type help dataex
clear
input long siteid str45 standardized_bank_name float year double it_budget
  1110931 "PNC BANK" 2011 231095
  1612514 "PNC BANK" 2011 121470
131106296 "PNC BANK" 2011 103732
139171356 "PNC BANK" 2011 142475
  1610215 "PNC BANK" 2011 259069
  1364277 "PNC BANK" 2011  34054
139219360 "PNC BANK" 2011  36203
121102690 "PNC BANK" 2011 147145
   562701 "PNC BANK" 2011 147145
  1201673 "PNC BANK" 2011 219348
139219285 "PNC BANK" 2011 142475
121102670 "PNC BANK" 2011 147145
  1991944 "PNC BANK" 2011  34054
136213797 "PNC BANK" 2011  13139
134158486 "PNC BANK" 2011  37388
136213796 "PNC BANK" 2011  13139
139142583 "PNC BANK" 2011 142486
  1219261 "PNC BANK" 2011 122156
109015244 "PNC BANK" 2011  37388
   561789 "PNC BANK" 2011 144899
121077855 "PNC BANK" 2011 130982
139142920 "PNC BANK" 2011  36203
131127152 "PNC BANK" 2011  36203
139136225 "PNC BANK" 2011 143505
118059480 "PNC BANK" 2011  35659
  1613308 "PNC BANK" 2011 133430
   517440 "PNC BANK" 2011 121846
  2663804 "PNC BANK" 2011 106030
131101785 "PNC BANK" 2011 467044
131127149 "PNC BANK" 2011 142486
  1617538 "PNC BANK" 2011 133430
  2174347 "PNC BANK" 2011 142486
  1611753 "PNC BANK" 2011 253711
  1216546 "PNC BANK" 2011 147145
  1112422 "PNC BANK" 2011 140349
139215908 "PNC BANK" 2011 869056
131118598 "PNC BANK" 2011 103288
  1213252 "PNC BANK" 2011 101992
139215033 "PNC BANK" 2011  36203
139114911 "PNC BANK" 2011 259839
 19397249 "PNC BANK" 2011  57511
114227847 "PNC BANK" 2011  34054
  2154716 "PNC BANK" 2011 103288
139155242 "PNC BANK" 2011  36203
136147273 "PNC BANK" 2011 134029
147115456 "PNC BANK" 2011  68284
136185587 "PNC BANK" 2011  34054
139144128 "PNC BANK" 2011 133430
139178250 "PNC BANK" 2011 142475
131080125 "PNC BANK" 2011  36203
  1623866 "PNC BANK" 2011 142475
  2170849 "PNC BANK" 2011 100173
144338695 "PNC BANK" 2011 101293
  1212811 "PNC BANK" 2011 138095
  2156847 "PNC BANK" 2011 142475
139169545 "PNC BANK" 2011 274444
108021329 "PNC BANK" 2011  37388
131096852 "PNC BANK" 2011 121470
139144346 "PNC BANK" 2011 218886
121053431 "PNC BANK" 2011 141111
139155241 "PNC BANK" 2011 137677
105538861 "PNC BANK" 2011  38636
139174377 "PNC BANK" 2011 142486
105541397 "PNC BANK" 2011  32081
139190442 "PNC BANK" 2011 142486
132031769 "PNC BANK" 2011  35761
131114261 "PNC BANK" 2011 143505
121065697 "PNC BANK" 2011 100581
  1206246 "PNC BANK" 2011 138095
  1222169 "PNC BANK" 2011 147145
133180716 "PNC BANK" 2011  36203
139155240 "PNC BANK" 2011 142486
131118660 "PNC BANK" 2011 142475
109019483 "PNC BANK" 2011  37388
139198228 "PNC BANK" 2011 142517
131110370 "PNC BANK" 2011 218886
  1113554 "PNC BANK" 2011 140349
  2166461 "PNC BANK" 2011 142475
  2166818 "PNC BANK" 2011 142475
139169735 "PNC BANK" 2011 142475
  1991896 "PNC BANK" 2011 114848
118059484 "PNC BANK" 2011  35659
  2188301 "PNC BANK" 2011 142486
133206889 "PNC BANK" 2011  36203
139143870 "PNC BANK" 2011 133430
  1630358 "PNC BANK" 2011 142486
139233439 "PNC BANK" 2011  36203
139234476 "PNC BANK" 2011 142486
  1634286 "PNC BANK" 2011 142486
136236249 "PNC BANK" 2011 134029
  2189054 "PNC BANK" 2011 137677
131124327 "PNC BANK" 2011 137677
118044046 "PNC BANK" 2011 111828
118056318 "PNC BANK" 2011 133606
105308028 "PNC BANK" 2011  32081
  1215045 "PNC BANK" 2011 147145
  2166472 "PNC BANK" 2011 121470
114252276 "PNC BANK" 2011 170597
  1611747 "PNC BANK" 2011 265016
139144286 "PNC BANK" 2011 148155
end

What I am trying to do is figure out which numbers in the column "it_budget" are "close" to each other. More specifically, I want to set a threshold that would be indicative of two numbers being close enough that I would call them the "same". However, I do not want to loop through the numbers and check the differences one-by-one. Is there a way to find which numbers in that column are within, let's say, 5% of each other without looping through each observation and mark them with a dummy called "close"?

[Clyde Schechter]

I don't see how creating a variable close will work here. It will tell you that a given bank has some other bank whose it budget is close to its own, but it won't tell you how many, nor which one(s). It makes more sense to link each bank to all the others that have an it budget within 5% of theirs. So I would think this is your best bet:

tempfile copy
save `copy'

gen ll = 0.95*it_budget
gen ul = 1.05*it_budget

rangejoin it_budget ll ul using `copy'
sort siteid it_budget_U

-rangejoin- is written by Robert Picard and is available from SSC. To use it, you must also install -rangestat-, by Robert Picard, Nick Cox, and Roberto Ferrer, also available from SSC.

That said, if you really want a dichotomous indicator with no other information:

gen ll = 0.95*it_budget
gen ul = 1.05*it_budget

rangestat (count) close = it_budget, interval(it_budget ll ul) excludeself
replace close == !!close