On 29/11/2007, Matthew Brown <morven(a)gmail.com> wrote: Populations determined from census results can just justifiably stated to full precision, since a census is an exact count, not an estimate. Of course, some people are always going to get missed out, but that's a matter of accuracy, not precision (which is what you meant, I think). Anyone announcing population figures not drawn directly from a census that state it more precisely than nearest 100 are just kidding themselves, IMHO, nearest thousand is probably more appropriate.

Which, I think, goes to show that one sig. fig. is insufficient precision. Also, it's interesting that you should mention that quote, since the very point I was making about census results was actually that they are *not* statistics. A statistic is a number calculated from a sample of a population, the result of a census is just the actual number of people (or whatever). I guess that means they're either lies or damned lies...

Thank you for your promotional commentary. Ec GerardM wrote:

Hoi, Thank you for using a bad example. Thank you for using political arguments that are completely beside the point. The question is, what is a "Commons database" good for. It is good for providing information in a set way in as many languages as possible. This does not mean machine translation, it means having information in one place make damned sure that the underlying information is correct and make sure that this information can be understood. The Wikimedia Foundation is about providing information. When there is information in info boxes that can be provided in a language like Tamil, Malayam, Kannada, Telugu, Gujarati, Georgian to name but a few with a completely different script, it makes sense to have at least that information available in info boxes in those languages. It makes perfect sense to have reputable information available in this way, a way that allows us to prove that at least the underlying information is correct. When we are able to provide information in info boxes, we do not "need" to provide the same information again. This however is common practice in any project; facts are repeated in text and extra information is provided to give it extra depth. The quality of machine translation is not the same for every language. This gets you first of all broken translations and slowly but surely acceptable language. When people complain about the big numbers of articles involved, they should appreciate that for a machine translation algorithm the number of articles is immaterial. For the development of machine translation it is actually a boon when there is a lot of language to process. So people that complain about large numbers of articles when machine is used just do not understand the process. Thanks, GerardM On Nov 29, 2007 7:29 PM, User 32X <wikipedia(a)32x.de> wrote: