Solution:
Very often, census is mission impossible - suppose we want to know the average age of a man, if we do census right now, we have to wait until they all die to know how long they actually lived! Isn't that rather useless?
Even in case where census is possible, it might give low return on high cost. If we knew the distribution, we already know how much error we are going to have in the sample statistics (such as the sample mean), and if the error is within a reasonable bound for the problem at hand, then it is good enough. Again, take the life span as an example, if we knew the life span distribution is normal, and we have already sampled a lot of human life span data, then there is no point getting even more, because we know the error is going to be small, we don't need to know whether the actual average life span is 77.124 or 77.123, we just don't care.
No comments:
Post a Comment