?

good question

Actually the board folk are trying to find a time when we can all watch the selection procedure together (in order to assure that no one simply chose someone they liked). It may make sense to post the list of eligible members (although maybe not, because that could involve posting real names here, which might be objectionable to some). If there is a way to post those names with handles (which may not be possible unless people use their AF handles as Facebook names), then we should do that. I believe the rules were that you had to like it by the stroke of midnight on Dec 31st 2011 (although I don’t remember the date exactly, because I’m not eligible so didn’t really care). Chad can correct me if I’m wrong here.

Ok, Chad and I just went through the procedure via screen sharing and have agreed on a lucky winner. There were 194 eligible participants. Each was given a random number via excel’s RAND() function. After randomizing the numbers, the list was then copied using Copy->Paste as->Values and sorted in order of random number. The person with the highest number in the list was chosen and is being contacted by Chad. In the event that they are not available to receive their prize within a reasonable length of time, the next highest number on the list will be selected, and so on. The initials of the current winner are: J.S.

Nerts! Those aren’t my initials

$500 with 194 contestants, that means that the expected value of clicking “Like” ultimately came out to around $2.50. Not bad for 1 second’s worth of work.

just for fun: why do you think it’s important for the random number that is assigned to each candidate to be sampled from the uniform distribution? for example, if you assign a random number to each candidate using say, the geometric distribution - certainly some numbers are more likely to occur than others, but this “bias” applies to all and does not favor a particular candidate. in other words, if the candidates are represented by independent, identically distributed random variables, they need not be uniform in order for your procedure to work - right?

That is true, provided that the order of the numbers is all that matters (which in this case it is) and the distribution is effectively continuous (non-discrete). If the random is used for a procedure that requires an interval or ratio measurement then the distribution matters very much. If numerical order is all that matters, then each point on the density function can be mapped to a cumulative distribution function and therefore to percentiles, and so the order can always be mapped to a uniform distribution. Again, that won’t work if the absolute difference or a ratio between numbers is needed for some calculation. In practice, we used the uniform distribution because it was easy (i.e. RAND() ), and because it obviously doesn’t prejudice the result. Using any other non-discrete (or quasi-non-discrete) method also doesn’t bias the result, but that fact is not obvious to many people, so we probably would have chosen a uniform distribution anyway just to avoid having to make the argument.

nerdgasm!

I understood each individual word, just not the order that you put them in

This is why I never passed quant

i still don’t understand why I didn’t win…

once the result comes out, it becomes a binary “win” or “no win”. distribution doesn’t matter anymore

…as long as the distribution is continuous.