Howto: "drop all following observations if all of them are positive"

[Jannic Cutura]

Dear All,
I (yet again) have a weird filtering problem, thatI would like to solve in Stata. Consider the following representative data:

 clear  
 ssc install dataex  
 input float (bondid date yield)
1    1    1.2
1    2    0.8
1    3    1.1
1    4    0.9
1    5    1
1    6    1.1
1    7    1.2
2    1    1
2    2    1.1
2    3    1.2
2    4    1.3
2    5    1.4
2    6    1.5
2    7    1.6
2    8    1.7
2    9    1.8
3    1    1.2
3    2    0.8
3    3    1.1
3    4    0.9
3    5    1
3    6    1.1
3    7    1.2
3    8    1.3
3    9    1.4
3    10    1.5
3    11    1.6
end

bys bondid (date): gen YieldChange =( yield-yield[_n-1])/yield[_n-1]

My Datastream yield data seems corrupted in that some bond yields start increasing exponentially towards maturity date (going into the gazillions). I assume the Datastream algorithm breaks down here. To avoid bias in the data I would like to drop all observations from where on the YieldChange is always positive for all following observations.

In the example above, this would mean:

• Bond 1: Drop 5-7
• Bond 2: Drop all
• Bond 3: Drop 5-11

Whenever problems become recursive, I am a bit lost in Stata. Can anyone point me to useful commands in this respect?
This is to be computed on 12,000 bonds and T= 4*365 so computationally efficiency may be important, but at first I would like to understand how to tackle something like this in Stata.

Thanks in advance!

Best,

[Nick Cox]

Always positive here seems to mean yield >1. (Like most of the most active members I do not deal with finance data.)

I suggest reversing time:

gsort bondid -date
by bondid : drop if sum(yield > 1) == 0

Compare the FAQ http://www.stata.com/support/faqs/da...t-occurrences/

[Jannic Cutura]

Nick Cox Thanks! Let me try this. What I meant was that if YieldChange is positive and for all subsequent obsevrations positive as well, then drop the observations. but your code should be easily adjustable -- let me try.