Creating a vector with unique observations from a variable in Stata - loops

Question

What I am mainly trying to do is to create a variable in which I can assign, within a stratum of my sample (defined by an 'id' variable, for instance), a name that is associated with the highest frequency (in the stratum) of this same name in another (string) variable. If tabulate* would work the way I need it to work, my code would run like this:
gen new_class_within_id=""
forvarlues i=1/80 {
tab class_var, matcell(x) if id==`i'
svmat x
sum x2
local name =x1 if x2==r(max)
replace new_class_within_id=`name' if id==`i'
}
That would be the general idea if tabulate would permit storing the unique observation names in a matrix -- the code might have some unintended errors too, of course. But while it does not seem to be possible using the above code, I thought that I could use mkmat if I would be able to store, in the loop, the unique observations inside a vector with some additional coding. Would that be possible? Also, is there an easier way to perform what I want to do?
*Firstly, I thought that using tabulate and extracting the results into a matrix would do the work that I need, but tabulate does not allow me to extract the names of the observations, just the frequencies. tabulate seemed nice because in its output it shows the unique observations of a variable in a column, but I could not find a way to extract those observations the way the output shows.

Answer 1

I think I understand your question, but maybe I don't. Some code:
clear
set more off
input ///
id str1 anothvar
1 a
1 a
1 a
1 b
1 m
2 c
2 c
2 m
2 a
2 z
end
list, sepby(id)
*-----
bysort id anothvar : gen count = _N
bysort id (count): gen newvar = anothvar[_N]
list, sepby(id)
More work needs to be done if you have missings and/or ties.