You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Only warn about best matches with different keys if the improvement in points is very significant. Say if the best match points is more than N std dev above random guessing. That is, if (1 - BINOM_CDF) is less than 0.0001 or something.
The text was updated successfully, but these errors were encountered:
For N_Q questions with N_A answers each, let F be the binomial CDF with parameters n = N_Q and p = 1/N_A.
The probability of a student answering at least k questions correctly by chance is F(k).
With N_s students taking N_e exams, the probability that some student answers at least k questions correctly when scored against an exam other that their own is:
p = 1 - F(k)^(N_s * (N_e - 1))
We should only print out warnings if this is value is below some threshold, say 0.01.
Only warn about best matches with different keys if the improvement in points is very significant. Say if the best match points is more than N std dev above random guessing. That is, if (1 - BINOM_CDF) is less than 0.0001 or something.
The text was updated successfully, but these errors were encountered: