How about attacking this problem directly:
bysort firm (year): gen num = _n
gen sum = 0
by firm: replace sum = income + income[1]/num[2] if num== 2
by firm: replace sum = income + income[1]/num[3] + income[2]/num[2] ///
if num == 3
by firm: replace sum = income + income[1]/num[4] + income[2]/num[3] + ///
income[3]/num[2] if num == 4
by firm: replace sum = income + income[1]/num[5] + income[2]/num[4] + ///
income[3]/num[3] + income[4]/num[2] if num == 5
by firm: replace sum = income + income[1]/num[6] + income[2]/num[5] + ///
income[3]/num[4] + income[4]/num[3] + income[5]/num[2] if num == 6
Hope this helps,
Scott
P.S. `=seq' is the same as seq[1] or 1
> -----Original Message-----
> From: [email protected] [mailto:owner-
> [email protected]] On Behalf Of Robin Luo
> Sent: Saturday, February 19, 2005 1:18 PM
> To: [email protected]
> Subject: st: question about calculating discounted sum
>
> I need to calculate discounted sum for a variable in a panel data. My
> data is like:
>
> FirmID year income
> 100 1980 50000
> 100 1981 51000
> 100 1982 54000
> 100 1983 60000
> 100 1984 59000
> 100 1985 62000
> 101 1970 18000
> 101 1971 18500
> 101 1972 20000
> 101 1973 24000
>
> Basically, the data is a panel data, grouped by "FirmID", with
> hundreds of frims. What I need to do is to calculate a moving
> discounted sum of income for each firm-year. That is, calculate a sum
> of income from a firm's first year to the current year, but each
> year's income will be discounted by that year's distance to the
> current year and the discount rate is the inverse of the distance
> "1/(current year - year +1)". For example, for firm 100 at 1982, this
> calculation would be like: 54000 + 51000*1/2 +50000*1/3, while for
> firm 100 at 1984 it would be like 59000 + 60000*1/2 + 54000*1/3 +
> 51000*1/4 + 50000*1/5. I tried the following program:
>
> by FirmID: gen seq=_n
>
> gen suminc=0
>
> quietly forval i=1/`=seq' {
> by FirmID: replace suminc = suminc + income[`i']/(`=seq' - `i'
> +1)
> }
>
> But problems are: 1) the loop seems not allowing "by" command; 2) the
> loop does not really launch, because even after I get rid of "by
> FirmID" the loop seems to go only one cycle instead of "_n" cycles.
>
> Many many thanks for any help!
>
> 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/
