Suppose I have a bunch of coins, all of which have some type of biased or unbiased weighting. For instance, a coin could come up heads 75% of the time, 50%, 25%, 5%, 0%, etc.
I am given a target heads bias. (e.g. 45%).
I want to classify the coins into either one of two classes: those coins with a heads bias above/at this target, and those coins with a heads bias below it.
I can flip each coin as many times as I want, but I want to find out the minimum number of flips I need per coin in order to do this classification. THAT IS:
(a) how many flips do I need so that the coin can be classified at all; and
(b) how many flips do I need to classify the coin to be above or below the target bias.
(a) and (b) are probably the same (I think).
It is ok if a confidence parameter is needed. e.g. I am 90% confident that this coin can be classified at all.
Someone said something about the inverse binomial or geometic distributions, but I am not sure if those are applicable here.
I am given a target heads bias. (e.g. 45%).
I want to classify the coins into either one of two classes: those coins with a heads bias above/at this target, and those coins with a heads bias below it.
I can flip each coin as many times as I want, but I want to find out the minimum number of flips I need per coin in order to do this classification. THAT IS:
(a) how many flips do I need so that the coin can be classified at all; and
(b) how many flips do I need to classify the coin to be above or below the target bias.
(a) and (b) are probably the same (I think).
It is ok if a confidence parameter is needed. e.g. I am 90% confident that this coin can be classified at all.
Someone said something about the inverse binomial or geometic distributions, but I am not sure if those are applicable here.
...is the expected probability (think percent...like x many times/ n number of tries) The Z score you should use here is 1.96 (which is derived from a 95% confidence level)
Facebook
Twitter