Refer to iteration number inside for-loop in Stata - stata

Question

Like many others, I often loop through variables in Stata, running some estimation command and then extracting the results to a variable created to hold them. This is simple when the variables are numbered sequentially or in some pattern (e.g. even numbers in a set). As an example:
sysuse auto
gen var1 = uniform()
gen var2 = uniform()
gen var3 = uniform()
*Create variables to hold results
gen str4 varname=""
gen results=.
*Loop through three variables
foreach i of numlist 1/3{
replace varname="var`i'" in `i'
sum var`i'
replace results=r(mean) in `i'
}
However, I often want to do something similar when the variables are not numeric and/or are not in an easy-to-handle order. Let's say I wanted to do the same thing for price, mpg, weight and length in the auto dataset. If we set up the for-loop as:
sysuse auto
gen str4 varname=""
gen results=.
foreach var of varlist price mpg weight length{
sum `var'
*Place values, in order, in rows?
}
then we need some way to understand that price is the first variable in the list, so its results should go in row 1 (or its name in row 1, or whatever we want to do).
This must be possible, but I would appreciate some suggestions. A clean/non-hackish way would be ideal, as I will be doing this a lot.

Answer 1

You can use a local counter that you start at 1 and increment at the end of each iteration:
sysuse auto, clear
gen varname=""
gen mean=.
local i=1
foreach var of varlist price mpg weight {
quietly sum `var'
replace mean = r(mean) in `i'
replace varname = "`var'" in `i'
local ++i
}

Answer 2

You could also do this. It's unlikely to seem as direct or simple as the standard technique explained by #Dimitriy V. Masterov, but it has its uses.
sysuse auto, clear
gen varname = ""
gen mean = .
local nvars : word count price mpg weight
tokenize "price mpg weight"
quietly forval j = 1/`nvars' {
sum ``j'', meanonly
replace mean = r(mean) in `j'
replace varname = "``j''" in `j'
}
The general points are
Words are separated by spaces, except that double quotation marks and compound double quotation marks bind tighter. Thus a, b and c are unsurprisingly the words in a b c but there are just two words in Stata "is great"
You can count how many objects you are looping over. It is the number of words in a string.
Applying tokenize to an argument string maps the separate words of that argument to local macros named 1, 2 and so forth. The nested macro references that is likely to imply are interpreted just as you would guess from elementary algebra: the innermost argument is evaluated first.
For more complicated problems, including the unpacking of a varlist, check out also unab.