climbing_vine
Well-known
N=30 is generally all that is needed for a p<.05. 😛
If, and only if, you're dealing with a truly random sample of a larger population, which isn't evident in the eBay case, especially for niche items. Possible, but not necessarily so.
Do you have any data to support the assertion that an overwhelming amount of ebay completed auction data are shill bids or unfulfilled sales? I agree that some are, but without knowledge of the magnitude it is difficult to discount the entire data pool.
Estimates are running as high as 90% in certain areas (like laptops), though of course eBay will not reveal the data to create a truly accurate picture. I found this paper interesting:
http://tinyurl.com/myu64f
First, because it's from four years ago, and we know the problem is much worse now; second, because they found that in certain conditions shill bidding actually seemed to artificially *reduce* winning bids. Which is still a distortion, but not the one I expected! I'd love to see an update on this sort of work.