I roll an n-sided die, n unknown, k times, and get p_1...p_k. Is there a way to estimate n from this?
Conversation
Replying to
I think just the obvious: take the mean of p_1...p_k, multiply by 2. Accuracy of the estimate increases with k.
1
1
Replying to
If you assign each face a value starting from 1 and ascending with each new glyph that appears the average of p_k will converge on (n+1)/2 assuming a fair die
1
4
This Tweet was deleted by the Tweet author. Learn more
Replying to
max(p_k) for sufficiently large k. Else, estimate assuming a binomial distribution with P(X=1) for max(p_k) i.e p=1/n. Mean would be k/n. If (k/max(p_k)) estimates the number of observations max(p_k) then n = ~max(p_k)
Replying to
If I understand your question correctly, yes. Use Beta updating to estimate the possible proportion of each side after each roll. I have a short tutorial on this.
1
Replying to
If not sequentially numbered, I wonder if you could look at it like the same birthday problem in reverse: if you have x pairs of people with the same birthday (pairs of rolls where p_A = p_B) in a room of k people, how many days are there in a year?
Replying to
I'd estimate that if 1 were to die N side the Unknown they are dead forever.
Show additional replies, including those that may contain offensive content
Show








