regression over different number of observations

[Philip Hiller]

Hi everyone

I would like to regress each variable separately on one independent variable. Furthermore, the first regression for each variable should be based on the observations 1 up to lets say 10, the second regression for each variable on the observations 11 up to 20 and so on. In a final step I would like to have all the coefficients listed in a table.

This is the code I thought may work, but it gives the error massage " 'i' invalid obs no". Could anyone tell me where the mistake is?

Any help would be very much appreciated!

clear all
set more off

sysuse auto

postutil clear

postfile bse str20 var float (beta1) using results, replace
foreach s of varlist mpg-trunk{
 forvalues i= 1 (1)64{
  quiet reg price `s' in ‘i’ / ‘i’ +10
  }
  matrix r = r(table)
  post bse ("`s'") (r[1,2])
}
postclose bse
use results, clear
list

[Mike Lacy]

The syntax for -in- requires something that evaluates to a number on both sides of the "/". You need to force evaluation of `i' + 10. You can put `i' + 10 into a local and reference that, or evaluate `i' + 10 on the fly:

forvalues ... {
   local ip10 = `i' + 10
   quiet reg ... in `i'/`ip10'
....

or

forvalues ... {
   quiet reg ... `i'/`=`i' + 10'
...

[Philip Hiller]

Now it works. Thank you very much!