Stata The Stata listserver
[Date Prev][Date Next][Thread Prev][Thread Next][Date index][Thread index]

st: very confused - how to implement a dynamic lagged variable


From   Robin Luo <rluosf@gmail.com>
To   statalist@hsphsun2.harvard.edu
Subject   st: very confused - how to implement a dynamic lagged variable
Date   Mon, 21 Feb 2005 01:50:55 -0800

I got really confused by this problem: I want to use a lagged variable
in my analysis but instead of lagging based on some pre-existing
value, this lag actually comes from certain calculation. My data is a
panel data (grouped by "firmid") with a structure like:

firmid	year	age_a	age_b	weight_a	weight_b
100	1960	10			
100	1961	11	2	.9	.1
100	1962	12			
100	1963	13	10	.7	.3
100	1964	14			
100	1965	15			
100	1966	16	5	.8	.2
101	1940	1			
101	1941	2	10	.6	.4
101	1942	3			

The goal is to calculate an adjusted age for each firm-year. In the
data, age_a is the natural age of the firm, while age_b comes from
certain event (say, an acquisition). Before there is any event, the
firm's adjusted age is simply the natural age (age_a). When there is
an event, the new adjusted age is the weighted sum of the adjusted age
and age_b (weight_a is the weight for adjusted age while weight_b for
age_b). When there is no event, the firm's adjusted age simply
increase by 1 for each year. I thus wrote the following program:


clear

input firmid year age_a age_b weight_a weight_b 
100 1960 10 . . .
100 1961 11 2 0.9 0.1
100 1962 12 . . .
100 1963 13 10 0.7 0.3
100 1964 14 . . .
100 1965 15 . . .
100 1966 16 5 0.8 0.2
101 1940 1 . . .
101 1941 2 10 0.6 0.4
101 1942 3 . . .
end

sort firmid year

gen age=0

by firmid: replace age=age_a if _n==1 & age_b==.
by firmid: replace age=(age_a*weight_a + age_b*weight_b) if _n==1 & age_b~=.

by firmid: replace age=(age[_n-1]+1)*weight_a + age_b*weight_b if
_n~=1 & age_b~=.
by firmid: replace age=(age[_n-1]+1) if _n~=1 & age_b==.


The result, however, came quite strange:

firmid	year	age_a	age_b	weight_a	weight_b	age
100	1960	10				10
100	1961	11	2	.9	.1	10.1
100	1962	12				11.1
100	1963	13	10	.7	.3	3.7
100	1964	14				4.7
100	1965	15				5.7
100	1966	16	5	.8	.2	1.8
101	1940	1				1
101	1941	2	10	.6	.4	5.2
101	1942	3				6.2

For the 4th and 7th observations (where "year" are 1963 and 1966
respectively), the correct ages should be about 11
([11.1+1]*0.7+10*0.3) and 6 ([5.7+1]*0.8+5*0.2). Obviously, instead of
getting the actual last "age" values, STATA took the very original
ones - the generated 0s. I could not figure it out and really
appreciate any help on this.

Thanks a lot!

Robin Luo
*
*   For searches and help try:
*   http://www.stata.com/support/faqs/res/findit.html
*   http://www.stata.com/support/statalist/faq
*   http://www.ats.ucla.edu/stat/stata/



© Copyright 1996–2014 StataCorp LP   |   Terms of use   |   Privacy   |   Contact us   |   What's new   |   Site index