Type Mismatch Error with tabcount

[Grace Domine]

The data contain several unique classes "class_id" (byte, values 1-20) with students "student_id" (byte, values 1-30, 1-29, 1-24, etc.).

Resp1 - Resp3 contain byte values of student_id.

Here is a snapshot of the data.

class_id	student_id	Resp1	Resp2	Resp3
1	1	9	7	5
1	2	14	4	22
1	3	1	.	22
1	28	30	1	1
1	29	2	22	4
1	30	9	.	18
2	1	3	15	13
2	2	17	25	1
2	3	3	17	9

I would like to count the frequency each student_id across observations of Resp1-Resp3 within each class_id.
My goal in using tabcount is to eventually create a variable that contains frequency of student_id for each observation of a student_id in a class_id.

Essentially, this command below, but across all Resp variables, with student_ids with zero occurrences included, missing excluded from display.
bysort student_id (class_id): tab PN1, matcell(matrix`Class')
svmat matrix`Class'

My thought is tabcount should be able to handle this. But when I run the following, I receive a type mismatch r(109) error. None of the variables are string. Could it be I'm misusing a macro? (I also don't quite understand how what arguments c() or v() are supposed to take.)

vl create Resp_list = (Resp1 Resp2 Resp3)
foreach var in $Resp_list{
di "Var is: `var'"
bysort class_id (student_id): tabcount `var', c(`var') zero
}

Any help is appreciated!

[Nick Cox]

tabcount is from SSC. You've tried to use a c() option, but such an option is for a series of equalities or inequalities and a variable is not such. So, you have got tabcount confused. I am the author of tabcount and I am no clearer on why you are trying to use it.

Please show what output you expect for

bysort class_id (student_id): tabcount Resp1, c(Resp1) zero

You want a series of tables it seems for each class_id and each of Resp1 Resp2 Resp3. How many tables will that be and what will you do with them?

[Grace Domine]

Thanks for the response.

I now understand that c() must take an inequality. I worked out the v() option instead.

I would like to use tabcount to answer the question: within each class_id, how many times does each student_id occur across Resp1-Resp3?

tabulate does not display frequencies for observations with 0 occurrences.

So I did this (below), which I can tweak in a larger forvar loop across the Resp1-Resp3 variables. I'll probably use the matrix option to store the results or push the matrix to a variable.

forvar in $Resp_list{
foreach level in `class_id'{
sum student_id if class_number==`l'
local id_max=r(max)
display `l'
tabcount `var' if class_id==`l', v(1/`id_max') zero
}
}

I've been able to create the desired output, more or less.

[Nick Cox]

Sounds good. I don't know (or have forgotten) where forvar comes from. Perhaps that's a typo and there seem to be others in the code.

I would rewrite this as

foreach var in Resp1 Resp2 Resp3 { 

    foreach level in `class_id' {
          display `level' 
          sum student_id if class_id == `level', meanonly 
          tabcount `var' if class_id == `level', v(1/`r(max)') zero
    }

}

But it seems that your real code worked "more or less".