variable (that exists) "not found" in loop - says Stata

[Anya Fedyk]

sorry for the weird thread title.

So here's what I write. It is a nested loop to find out whether some change in a variable bla corresponds approximately to a change in the variable bla for another observation within the same group:


forvalues group = 1/2 {

su position if group== `group' , meanonly

forvalues from = 1/`r(max)' {
forvalues to = 1/`r(max)' {

bysort group: replace from_to = "`from'+"to"+`to'" if d_bla[`_from'] < upperbound[`_to'] & d_bla[`_from'] > lowerbound[`_to'] & group==`group'

}
}

}


and I get the following error:


d_bla not found
r(111);



however, the variable exists. can anybody tell me what the error is?
I know this will not be the most efficient way of programming, it's just the first solution I came up with...
thank you very much!

[Rich Goldstein]

show us the result of "d d_bla"

also, why are you putting underscores before "from" and "to" in your bysort statement? they don't belong

[Nick Cox]

On this evidence you are referring to local macros _from and _to that you have not defined. However, what Stata is complaining about is that it doesn't know what d_bla is. Rich is right that if it were a variable describe would show it, but I doubt you would get that error message if it were a variable. Similar, but not identical, comments if it were a scalar.

In short, we can't advise here easily without knowing what everything mentioned in this code is supposed to be.

[Anya Fedyk]

thank you for your answers. I appreciate your help, unfortunately I could not access this until now.

1) desc yields:

storage display value
variable name type format label variable label
-------------------------------------------------------------------------------------
d_bla float %9.0g

2) I put underscores because I wanted to refer to observation `from' or `to' and thought I could achieve this with underscores, yielding results as in "_n". Is this not the case?


/update: it works now without the underscores. It just takes VERY much time from what I can estimate. I will get back to you if it does not do what I wanted it to do. Thank you again!

[Nick Cox]

You were clearly right that d_bla exists as a variable. As you haven't defined local macros _from or _to (and as they do not have in-built roles) your code will look to Stata like

forvalues group = 1/2 {
  su position if group== `group' , meanonly
  forvalues from = 1/`r(max)' {
    forvalues to = 1/`r(max)' {
      bysort group: replace from_to = "`from'+"to"+`to'" if d_bla[] < upperbound[] & d_bla[] lowerbound[] & group==`group'
    }
  }
}

My guess about the error message you got was wrong. But my next guess is that Stata was misled by the [] to thinking that you were trying to access a matrix with that name.

Regardless of that, what you are trying to do with this code is quite unclear to me. I'd back up and explain what you want, with example data and example results.