sistawendy (
sistawendy) wrote2010-11-11 10:37 pm
![[personal profile]](https://www.dreamwidth.org/img/silk/identity/user.png)
Entry tags:
Nun solves grade inflation.
Almost two years ago I took a stab at solving the grade inflation problem. Basically, my earlier scheme boils down to a ranking system, but it's missing a couple of things: slack for teachers, and a guarantee that how teachers use that slack won't screw students. Here goes.
Let nk be the number of students in class k where k is in [1..c], and let there be ranks i in [1..mk] with bi,k students in each, where higher i means the student has done better, such that for all i and k bi,k > 0, and
Sum[i=1..m](bi,k) = nk
Administrators can impose constraints like for all i and k, mk >= 3 and max(bi,k / mk) <= 0.7.
Here's where it gets interesting. Define
wk = Sum[i=1..mk](ibi,k)
Let rj,k where j is in [1..nk] be the rank to which student j belongs. The score s(j,k) for a student in rank rj,k is
s(j,k) = 2^(rj,k/wk) - 1
giving you a score strictly between 0 and 1 and guaranteeing that for all k
Product[j=1..nk](s(j,k)+1) = 2
Student j's cumulative score across all classes is then
S(j)=[Product[k=1..c](s(j,k))](1/c)
which will also be strictly between 0 and 1. The higher S(j) is, the better a student j has been, say his teachers. Assign letters to specific, non-overlapping ranges between 0 and 1 if you must, but never, ever change them.
Teachers, especially those who aren't numerically inclined, may squawk, but I could write an Excel spreadsheet to compute s(j,k) and check at least some of the constraints.
Please forgive my bastard notation. Next: a closed-form solution to the Arab-Israeli conflict.
ETA: The underscore is to indicate a subscript.
ETAA: b_i,k is b with two subscripts, i and k. Curse you,
neuro42!
Let nk be the number of students in class k where k is in [1..c], and let there be ranks i in [1..mk] with bi,k students in each, where higher i means the student has done better, such that for all i and k bi,k > 0, and
Administrators can impose constraints like for all i and k, mk >= 3 and max(bi,k / mk) <= 0.7.
Here's where it gets interesting. Define
Let rj,k where j is in [1..nk] be the rank to which student j belongs. The score s(j,k) for a student in rank rj,k is
giving you a score strictly between 0 and 1 and guaranteeing that for all k
Student j's cumulative score across all classes is then
which will also be strictly between 0 and 1. The higher S(j) is, the better a student j has been, say his teachers. Assign letters to specific, non-overlapping ranges between 0 and 1 if you must, but never, ever change them.
Teachers, especially those who aren't numerically inclined, may squawk, but I could write an Excel spreadsheet to compute s(j,k) and check at least some of the constraints.
Please forgive my bastard notation. Next: a closed-form solution to the Arab-Israeli conflict.
ETAA: b_i,k is b with two subscripts, i and k.
![[livejournal.com profile]](https://www.dreamwidth.org/img/external/lj-userinfo.gif)
no subject
no subject
no subject
no subject
no subject
no subject