tag:blogger.com,1999:blog-3366935554564939610.post2273449026313693722..comments2020-08-29T12:11:06.773-04:00Comments on Data Miners Blog: The Agent Problem: Sampling From A Finite PopulationMichael J. A. Berryhttp://www.blogger.com/profile/06077102677195066016noreply@blogger.comBlogger2125tag:blogger.com,1999:blog-3366935554564939610.post-18515346512134161182008-09-13T12:23:00.000-04:002008-09-13T12:23:00.000-04:00I haven't actually found any good sources for this...I haven't actually found any good sources for this, although you can search "finite population correction factor on the web". It is also mentioned on the wikipedia page for margin of error (http://en.wikipedia.org/wiki/Margin_of_Error) under the heading "Population Adjustment".<BR/><BR/>When I faced this problem, I asked several statisticians. Surpringly, the one who was able to answer had a masters in statistics and had specialized in surveys. Perhaps this is not so surprising, but several other PhDs did not know the answer.<BR/><BR/>I haven't yet found a proof of the finite correction factor.<BR/><BR/>As for your second question, I need the adjustment so the standard error is not 0. The same situation arises when assuming an infinite population, so it has become a habit over time.<BR/><BR/>Looking around the web,I came across references to the Wilson Estimate, which adds two successes and failures to the observed data before estimating the standard error. In essence, this is starting with a prior of 50% and giving it a weight of two observations.<BR/><BR/>My technique of subtracting half a success or failure actually <BR/>creates very similar estimates, when the observed population has a p value of 0% or 100%.Gordon S. Linoffhttps://www.blogger.com/profile/02341184075032239786noreply@blogger.comtag:blogger.com,1999:blog-3366935554564939610.post-56575711529080090132008-09-09T10:52:00.000-04:002008-09-09T10:52:00.000-04:00Thanks for the clear, understandable explanation o...Thanks for the clear, understandable explanation of finite populations and the fpc. Do you have suggestions for further reading? I do work in which I need to analyze effects of small populations.<BR/><BR/>Why do you "add or subtract 0.5 from the proportion to calculate the standard error" hwen the proportion is 0 or 1? Is there a statistical reason or is it an arbitrary correction?Anonymoushttps://www.blogger.com/profile/01432097266338789860noreply@blogger.com