# Generating - transformation of a time series variable

Hi all,

I have the following problem: I want to generate a new time series variable which measures the accumulated (the sum) of an existing variable "X". The new variable is a sum inside a given moving time window with a constant width of let say 40 years. More precisely - the new variable is generating in the following way: Let the sum of the first 40 observations from time point t = 1 to t = 40 represents the first observation, the second observation is the sum of the observations from t = 2 to time point t = 41, the third observation is the sum of observations from t = 3 to t = 42 etc. The constant time window of 40 years is moving by one step at the time. It's like a moving average but WITHOUT dividing by the number of observations. I hope one of you Shazam experts could look into the problem - it would be of great help.

Regards, P.

edit retag close merge delete

Sort by » oldest newest most voted
* Program to create moving sums of a variable
sample 1 10
* generate some simple data to make it easy to check
genr x=time(0)
* initialize the moving sum variable
genr msum=0
* add up the number of lags (this example does current plus last 2)
* if you dont like output, you can SET NODOECHO after it works
do #=0,2
genr msum=msum+lag(x,#)
endo
* print it out, the first few observation will be garbage
print x msum
stop

more

Ken - Thanks a lot for your help - it function perfectly! :)

( 2012-06-27 01:55:56 +0000 )edit

## Stats

Seen: 417 times

Last updated: Jun 26 '12