Here's the problem:
S(p) = SUM_{k>0} U_k * (1-p)^k
N := SUM_{k>0} U_k is fixed: it's the number of users
M := SUM_{k>0} k * U_k is fixed: it's the number of pages
If the U_k's are not known, but we know M/N, how can we approximate the
ratio S(p)/N? lower and upper bounds? median? Typically, p = 20%.
U_k is the number of users turning exactly k pages. The model is described
at http://www.datashaping.com/internet.shtml .
|