# st: looping questions

 From Kit Baum To statalist@hsphsun2.harvard.edu Subject st: looping questions Date Tue, 9 Aug 2005 08:53:52 -0400

> How can I write a fundction that calculates the change between successive
> observations of 10 elements of this variable? Ie. Assume the data is
> sorted by date, decending. The first observation is
> (Var1[_n] -Var1[_n+10])/Var[_n+10], the second observation is
> (Var1[_n+2]-Var1[_n+11])/Var1[_n+11], and so on...Is this an instance
> where I can use the -rolling- command? If so, how?

and R.E. replied

local i=10
while `i'<_N { //"_N" is the number of observations
replace myvar = (Var[_n]-Var[_n+`i'])/Var[_n+`i']) if _n=(`i'-9)
local i=`i'+1
}

The original formula is very strange; normally we calculate a change as the difference between now and a while ago, not the other way around. That is, to calculate a year-over-year price change from monthly data, I would do

infl_t = ( P_t - P_{t-12} ) / P_{t-12}

But presuming Kelly really wants what she says, and that these are time series data (or could be tsset as such) a more Stataish way to do that would be something like

webuse wpi1
g wpid = (wpi - F10.wpi)/F10.wpi

which will avoid the explicit "while" loop (always a good idea). In fact one would never need the explicit loop; you could do

g wpid = (wpi - wpi[_n+10] ) / wpi[_n+10]

but I would much rather use the time series operators. They don't make mistakes.

If you want to loop this calculation, just put the variable names in a local macro and use a foreach loop. Or, if they are numbered x1- x100, use a forvalues loop, e.g.

forvalues v=1/100 {
g wpi`v' = etc.

Kit Baum, Boston College Economics
http://ideas.repec.org/e/pba1.html

*
* 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/